Merge http://fbdev.bkbits.net/fbdev-2.5

into home.transmeta.com:/home/torvalds/v2.5/linux

Documentation/DocBook/kernel-api.tmpl

@@ -299,6 +299,8 @@ KAO -->
     EHCI, OHCI, or UHCI.
     </para>
 !Edrivers/usb/core/hcd.c
+!Edrivers/usb/core/hcd-pci.c
+!Edrivers/usb/core/buffer.c
     </sect1>
 
   </chapter>

arch/i386/kernel/apm.c

@@ -214,6 +214,7 @@
 #include <linux/sched.h>
 #include <linux/pm.h>
 #include <linux/kernel.h>
+#include <linux/smp.h>
 #include <linux/smp_lock.h>
 
 #include <asm/system.h>
@@ -419,6 +420,7 @@ static DECLARE_WAIT_QUEUE_HEAD(apm_waitqueue);
 static DECLARE_WAIT_QUEUE_HEAD(apm_suspend_waitqueue);
 static struct apm_user *	user_list;
 static spinlock_t		user_list_lock = SPIN_LOCK_UNLOCKED;
+static struct desc_struct	bad_bios_desc = { 0, 0x00409200 };
 
 static char			driver_version[] = "1.16";	/* no spaces */
 
@@ -568,7 +570,12 @@ static u8 apm_bios_call(u32 func, u32 ebx_in, u32 ecx_in,
 	u32 *eax, u32 *ebx, u32 *ecx, u32 *edx, u32 *esi)
 {
 	APM_DECL_SEGS
-	unsigned long	flags;
+	unsigned long		flags;
+	int			cpu = smp_processor_id();
+	struct desc_struct	save_desc_40;
+
+	save_desc_40 = cpu_gdt_table[cpu][0x40 / 8];
+	cpu_gdt_table[cpu][0x40 / 8] = bad_bios_desc;
 
 	local_save_flags(flags);
 	APM_DO_CLI;
@@ -591,6 +598,7 @@ static u8 apm_bios_call(u32 func, u32 ebx_in, u32 ecx_in,
 		: "memory", "cc");
 	APM_DO_RESTORE_SEGS;
 	local_irq_restore(flags);
+	cpu_gdt_table[cpu][0x40 / 8] = save_desc_40;
 	return *eax & 0xff;
 }
 
@@ -613,6 +621,11 @@ static u8 apm_bios_call_simple(u32 func, u32 ebx_in, u32 ecx_in, u32 *eax)
 	u8		error;
 	APM_DECL_SEGS
 	unsigned long	flags;
+	int			cpu = smp_processor_id();
+	struct desc_struct	save_desc_40;
+
+	save_desc_40 = cpu_gdt_table[cpu][0x40 / 8];
+	cpu_gdt_table[cpu][0x40 / 8] = bad_bios_desc;
 
 	local_save_flags(flags);
 	APM_DO_CLI;
@@ -639,6 +652,7 @@ static u8 apm_bios_call_simple(u32 func, u32 ebx_in, u32 ecx_in, u32 *eax)
 	}
 	APM_DO_RESTORE_SEGS;
 	local_irq_restore(flags);
+	cpu_gdt_table[smp_processor_id()][0x40 / 8] = save_desc_40;
 	return error;
 }
 
@@ -931,9 +945,9 @@ static void handle_poweroff (int key, struct pt_regs *pt_regs,
 }
 
 static struct sysrq_key_op	sysrq_poweroff_op = {
-	handler:        handle_poweroff,
-	help_msg:       "Off",
-	action_msg:     "Power Off\n"
+	.handler        = handle_poweroff,
+	.help_msg       = "Off",
+	.action_msg     = "Power Off\n"
 };
 
 
@@ -1826,12 +1840,12 @@ __setup("apm=", apm_setup);
 #endif
 
 static struct file_operations apm_bios_fops = {
-	owner:		THIS_MODULE,
-	read:		do_read,
-	poll:		do_poll,
-	ioctl:		do_ioctl,
-	open:		do_open,
-	release:	do_release,
+	.owner		= THIS_MODULE,
+	.read		= do_read,
+	.poll		= do_poll,
+	.ioctl		= do_ioctl,
+	.open		= do_open,
+	.release	= do_release,
 };
 
 static struct miscdevice apm_device = {
@@ -1927,13 +1941,13 @@ static int __init apm_init(void)
 	 * NOTE: on SMP we call into the APM BIOS only on CPU#0, so it's
 	 * enough to modify CPU#0's GDT.
 	 */
-	for (i = 0; i < NR_CPUS; i++) {
-		set_base(cpu_gdt_table[i][APM_40 >> 3],
-			 __va((unsigned long)0x40 << 4));
-		_set_limit((char *)&cpu_gdt_table[i][APM_40 >> 3], 4095 - (0x40 << 4));
+	set_base(bad_bios_desc, __va((unsigned long)0x40 << 4));
+	_set_limit((char *)&bad_bios_desc, 4095 - (0x40 << 4));
+
+	apm_bios_entry.offset = apm_info.bios.offset;
+	apm_bios_entry.segment = APM_CS;
 
-		apm_bios_entry.offset = apm_info.bios.offset;
-		apm_bios_entry.segment = APM_CS;
+	for (i = 0; i < NR_CPUS; i++) {
 		set_base(cpu_gdt_table[i][APM_CS >> 3],
 			 __va((unsigned long)apm_info.bios.cseg << 4));
 		set_base(cpu_gdt_table[i][APM_CS_16 >> 3],

arch/i386/kernel/bluesmoke.c

@@ -307,7 +307,7 @@ static void do_mce_timer(void *data)
 } 
 
 static struct tq_struct mce_task = { 
-	routine: do_mce_timer	
+	.routine = do_mce_timer	
 };
 
 static void mce_timerfunc (unsigned long data)

arch/i386/kernel/cpu/amd.c

@@ -189,8 +189,8 @@ static unsigned int amd_size_cache(struct cpuinfo_x86 * c, unsigned int size)
 }
 
 static struct cpu_dev amd_cpu_dev __initdata = {
-	c_vendor:	"AMD",
-	c_ident: 	{ "AuthenticAMD" },
+	.c_vendor	= "AMD",
+	.c_ident 	= { "AuthenticAMD" },
 	c_models: {
 		{ X86_VENDOR_AMD,	4,
 		  {
@@ -203,9 +203,9 @@ static struct cpu_dev amd_cpu_dev __initdata = {
 		  }
 		},
 	},
-	c_init:		init_amd,
-	c_identify:	amd_identify,
-	c_size_cache:	amd_size_cache,
+	.c_init		= init_amd,
+	.c_identify	= amd_identify,
+	.c_size_cache	= amd_size_cache,
 };
 
 int __init amd_init_cpu(void)

arch/i386/kernel/cpu/centaur.c

@@ -411,10 +411,10 @@ static unsigned int centaur_size_cache(struct cpuinfo_x86 * c, unsigned int size
 }
 
 static struct cpu_dev centaur_cpu_dev __initdata = {
-	c_vendor:	"Centaur",
-	c_ident:	{ "CentaurHauls" },
-	c_init:		init_centaur,
-	c_size_cache:	centaur_size_cache,
+	.c_vendor	= "Centaur",
+	.c_ident	= { "CentaurHauls" },
+	.c_init		= init_centaur,
+	.c_size_cache	= centaur_size_cache,
 };
 
 int __init centaur_init_cpu(void)

arch/i386/kernel/cpu/common.c

@@ -31,7 +31,7 @@ static void default_init(struct cpuinfo_x86 * c)
 }
 
 static struct cpu_dev default_cpu = {
-	c_init:	default_init,
+	.c_init	= default_init,
 };
 static struct cpu_dev * this_cpu = &default_cpu;
 

arch/i386/kernel/cpu/cyrix.c

@@ -322,10 +322,10 @@ static void cyrix_identify(struct cpuinfo_x86 * c)
 }
 
 static struct cpu_dev cyrix_cpu_dev __initdata = {
-	c_vendor:	"Cyrix",
-	c_ident: 	{ "CyrixInstead" },
-	c_init:		init_cyrix,
-	c_identify:	cyrix_identify,
+	.c_vendor	= "Cyrix",
+	.c_ident 	= { "CyrixInstead" },
+	.c_init		= init_cyrix,
+	.c_identify	= cyrix_identify,
 };
 
 int __init cyrix_init_cpu(void)
@@ -337,10 +337,10 @@ int __init cyrix_init_cpu(void)
 //early_arch_initcall(cyrix_init_cpu);
 
 static struct cpu_dev nsc_cpu_dev __initdata = {
-	c_vendor:	"NSC",
-	c_ident: 	{ "Geode by NSC" },
-	c_init:		init_cyrix,
-	c_identify:	generic_identify,
+	.c_vendor	= "NSC",
+	.c_ident 	= { "Geode by NSC" },
+	.c_init		= init_cyrix,
+	.c_identify	= generic_identify,
 };
 
 int __init nsc_init_cpu(void)

arch/i386/kernel/cpu/intel.c

@@ -327,8 +327,8 @@ static unsigned int intel_size_cache(struct cpuinfo_x86 * c, unsigned int size)
 }
 
 static struct cpu_dev intel_cpu_dev __initdata = {
-	c_vendor:	"Intel",
-	c_ident: 	{ "GenuineIntel" },
+	.c_vendor	= "Intel",
+	.c_ident 	= { "GenuineIntel" },
 	c_models: {
 		{ X86_VENDOR_INTEL,	4,
 		  { 
@@ -375,9 +375,9 @@ static struct cpu_dev intel_cpu_dev __initdata = {
 		  }
 		},
 	},
-	c_init:		init_intel,
-	c_identify:	generic_identify,
-	c_size_cache:	intel_size_cache,
+	.c_init		= init_intel,
+	.c_identify	= generic_identify,
+	.c_size_cache	= intel_size_cache,
 };
 
 __init int intel_cpu_init(void)

arch/i386/kernel/cpu/nexgen.c

@@ -42,13 +42,13 @@ static void nexgen_identify(struct cpuinfo_x86 * c)
 }
 
 static struct cpu_dev nexgen_cpu_dev __initdata = {
-	c_vendor:	"Nexgen",
-	c_ident:	{ "NexGenDriven" },
+	.c_vendor	= "Nexgen",
+	.c_ident	= { "NexGenDriven" },
 	c_models: {
 		{ X86_VENDOR_NEXGEN,5, { [1] "Nx586" } },
 	},
-	c_init:		init_nexgen,
-	c_identify:	nexgen_identify,
+	.c_init		= init_nexgen,
+	.c_identify	= nexgen_identify,
 };
 
 int __init nexgen_init_cpu(void)

arch/i386/kernel/cpu/proc.c

@@ -119,8 +119,8 @@ static void c_stop(struct seq_file *m, void *v)
 {
 }
 struct seq_operations cpuinfo_op = {
-	start:	c_start,
-	next:	c_next,
-	stop:	c_stop,
-	show:	show_cpuinfo,
+	.start	= c_start,
+	.next	= c_next,
+	.stop	= c_stop,
+	.show	= show_cpuinfo,
 };

arch/i386/kernel/cpu/rise.c

@@ -29,8 +29,8 @@ static void __init init_rise(struct cpuinfo_x86 *c)
 }
 
 static struct cpu_dev rise_cpu_dev __initdata = {
-	c_vendor:	"Rise",
-	c_ident:	{ "RiseRiseRise" },
+	.c_vendor	= "Rise",
+	.c_ident	= { "RiseRiseRise" },
 	c_models: {
 		{ X86_VENDOR_RISE,	5,
 		  { 
@@ -41,7 +41,7 @@ static struct cpu_dev rise_cpu_dev __initdata = {
 		  }
 		},
 	},
-	c_init:		init_rise,
+	.c_init		= init_rise,
 };
 
 int __init rise_init_cpu(void)

arch/i386/kernel/cpu/transmeta.c

@@ -80,10 +80,10 @@ static void transmeta_identify(struct cpuinfo_x86 * c)
 }
 
 static struct cpu_dev transmeta_cpu_dev __initdata = {
-	c_vendor:	"Transmeta",
-	c_ident:	{ "GenuineTMx86", "TransmetaCPU" },
-	c_init:		init_transmeta,
-	c_identify:	transmeta_identify,
+	.c_vendor	= "Transmeta",
+	.c_ident	= { "GenuineTMx86", "TransmetaCPU" },
+	.c_init		= init_transmeta,
+	.c_identify	= transmeta_identify,
 };
 
 int __init transmeta_init_cpu(void)

arch/i386/kernel/cpu/umc.c

@@ -11,8 +11,8 @@ static void __init init_umc(struct cpuinfo_x86 * c)
 }
 
 static struct cpu_dev umc_cpu_dev __initdata = {
-	c_vendor:	"UMC",
-	c_ident: 	{ "UMC UMC UMC" },
+	.c_vendor	= "UMC",
+	.c_ident 	= { "UMC UMC UMC" },
 	c_models: {
 		{ X86_VENDOR_UMC,	4,
 		  { 
@@ -21,7 +21,7 @@ static struct cpu_dev umc_cpu_dev __initdata = {
 		  }
 		},
 	},
-	c_init:		init_umc,
+	.c_init		= init_umc,
 };
 
 int __init umc_init_cpu(void)

arch/i386/kernel/cpuid.c

@@ -146,10 +146,10 @@ static int cpuid_open(struct inode *inode, struct file *file)
  * File operations we support
  */
 static struct file_operations cpuid_fops = {
-  owner:	THIS_MODULE,
-  llseek:	cpuid_seek,
-  read:		cpuid_read,
-  open:		cpuid_open,
+  .owner	= THIS_MODULE,
+  .llseek	= cpuid_seek,
+  .read		= cpuid_read,
+  .open		= cpuid_open,
 };
 
 int __init cpuid_init(void)

arch/i386/kernel/i8259.c

@@ -247,13 +247,13 @@ static int i8259A_resume(struct device *dev, u32 level)
 }
 
 static struct device_driver driver_i8259A = {
-	resume:		i8259A_resume,
+	.resume		= i8259A_resume,
 };
 
 static struct device device_i8259A = {
-	name:	       	"i8259A",
-	bus_id:		"0020",
-	driver:		&driver_i8259A,
+	.name	       	= "i8259A",
+	.bus_id		= "0020",
+	.driver		= &driver_i8259A,
 };
 
 static int __init init_8259A_devicefs(void)

arch/i386/kernel/microcode.c

@@ -109,17 +109,17 @@ static unsigned int mc_fsize;       /* file size of /dev/cpu/microcode */
 
 /* we share file_operations between misc and devfs mechanisms */
 static struct file_operations microcode_fops = {
-	owner:		THIS_MODULE,
-	read:		microcode_read,
-	write:		microcode_write,
-	ioctl:		microcode_ioctl,
-	open:		microcode_open,
+	.owner		= THIS_MODULE,
+	.read		= microcode_read,
+	.write		= microcode_write,
+	.ioctl		= microcode_ioctl,
+	.open		= microcode_open,
 };
 
 static struct miscdevice microcode_dev = {
-	minor: MICROCODE_MINOR,
-	name:	"microcode",
-	fops:	&microcode_fops,
+	.minor = MICROCODE_MINOR,
+	.name	= "microcode",
+	.fops	= &microcode_fops,
 };
 
 static devfs_handle_t devfs_handle;

arch/i386/kernel/msr.c

@@ -246,11 +246,11 @@ static int msr_open(struct inode *inode, struct file *file)
  * File operations we support
  */
 static struct file_operations msr_fops = {
-  owner:	THIS_MODULE,
-  llseek:	msr_seek,
-  read:		msr_read,
-  write:	msr_write,
-  open:		msr_open,
+  .owner	= THIS_MODULE,
+  .llseek	= msr_seek,
+  .read		= msr_read,
+  .write	= msr_write,
+  .open		= msr_open,
 };
 
 int __init msr_init(void)

arch/i386/kernel/mtrr.c

@@ -1836,11 +1836,11 @@ static int mtrr_close (struct inode *ino, struct file *file)
 
 static struct file_operations mtrr_fops =
 {
-    owner:	THIS_MODULE,
-    read:	mtrr_read,
-    write:	mtrr_write,
-    ioctl:	mtrr_ioctl,
-    release:	mtrr_close,
+    .owner	= THIS_MODULE,
+    .read	= mtrr_read,
+    .write	= mtrr_write,
+    .ioctl	= mtrr_ioctl,
+    .release	= mtrr_close,
 };
 
 #  ifdef CONFIG_PROC_FS

arch/i386/kernel/time.c

@@ -639,8 +639,8 @@ static unsigned long __init calibrate_tsc(void)
 }
 
 static struct device device_i8253 = {
-	name:	       	"i8253",
-	bus_id:		"0040",
+	.name	       	= "i8253",
+	.bus_id		= "0040",
 };
 
 static int time_init_driverfs(void)

arch/i386/mm/ioremap.c

@@ -159,7 +159,7 @@ void * __ioremap(unsigned long phys_addr, unsigned long size, unsigned long flag
 	area->phys_addr = phys_addr;
 	addr = area->addr;
 	if (remap_area_pages(VMALLOC_VMADDR(addr), phys_addr, size, flags)) {
-		vfree(addr);
+		vunmap(addr);
 		return NULL;
 	}
 	return (void *) (offset + (char *)addr);
@@ -215,13 +215,13 @@ void iounmap(void *addr)
 	struct vm_struct *p;
 	if (addr <= high_memory) 
 		return; 
-	p = remove_kernel_area((void *) (PAGE_MASK & (unsigned long) addr));
+	p = remove_vm_area((void *) (PAGE_MASK & (unsigned long) addr));
 	if (!p) { 
 		printk("__iounmap: bad address %p\n", addr);
 		return;
 	} 
 
-	vmfree_area_pages(VMALLOC_VMADDR(p->addr), p->size);	
+	unmap_vm_area(p);
 	if (p->flags && p->phys_addr < virt_to_phys(high_memory)) { 
 		change_page_attr(virt_to_page(__va(p->phys_addr)),
 				 p->size >> PAGE_SHIFT,

arch/sparc64/mm/modutil.c

@@ -4,6 +4,8 @@
  *  Copyright (C) 1997,1998 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
  *  Based upon code written by Linus Torvalds and others.
  */
+
+#warning "major untested changes to this file  --hch (2002/08/05)"
  
 #include <linux/slab.h>
 #include <linux/vmalloc.h>
@@ -16,6 +18,7 @@ static struct vm_struct * modvmlist = NULL;
 void module_unmap (void * addr)
 {
 	struct vm_struct **p, *tmp;
+	int i;
 
 	if (!addr)
 		return;
@@ -23,21 +26,38 @@ void module_unmap (void * addr)
 		printk("Trying to unmap module with bad address (%p)\n", addr);
 		return;
 	}
+
 	for (p = &modvmlist ; (tmp = *p) ; p = &tmp->next) {
 		if (tmp->addr == addr) {
 			*p = tmp->next;
-			vmfree_area_pages(VMALLOC_VMADDR(tmp->addr), tmp->size);
-			kfree(tmp);
-			return;
 		}
 	}
 	printk("Trying to unmap nonexistent module vm area (%p)\n", addr);
+	return;
+
+found:
+
+	unmap_vm_area(tmp);
+	
+	for (i = 0; i < tmp->nr_pages; i++) {
+		if (unlikely(!tmp->pages[i]))
+			BUG();
+		__free_page(tmp->pages[i]);
+	}
+
+	kfree(tmp->pages);
+	kfree(tmp);
 }
 
+
 void * module_map (unsigned long size)
 {
-	void * addr;
 	struct vm_struct **p, *tmp, *area;
+	struct vm_struct *area;
+	struct page **pages;
+	void * addr;
+	unsigned int nr_pages, array_size, i;
+
 
 	size = PAGE_ALIGN(size);
 	if (!size || size > MODULES_LEN) return NULL;
@@ -55,11 +75,32 @@ void * module_map (unsigned long size)
 	area->size = size + PAGE_SIZE;
 	area->addr = addr;
 	area->next = *p;
+	area->pages = NULL;
+	area->nr_pages = 0;
+	area->phys_addr = 0;
 	*p = area;
 
-	if (vmalloc_area_pages(VMALLOC_VMADDR(addr), size, GFP_KERNEL, PAGE_KERNEL)) {
-		module_unmap(addr);
+	nr_pages = (size+PAGE_SIZE) >> PAGE_SHIFT;
+	array_size = (nr_pages * sizeof(struct page *));
+
+	area->nr_pages = nr_pages;
+	area->pages = pages = kmalloc(array_size, (gfp_mask & ~__GFP_HIGHMEM));
+	if (!area->pages)
 		return NULL;
+	memset(area->pages, 0, array_size);
+
+	for (i = 0; i < area->nr_pages; i++) {
+		area->pages[i] = alloc_page(gfp_mask);
+		if (unlikely(!area->pages[i]))
+			goto fail;
 	}
-	return addr;
+	
+	if (map_vm_area(area, prot, &pages))
+		goto fail;
+	return area->addr;
+
+fail:
+	vfree(area->addr);
+	return NULL;
+}
 }

drivers/block/cciss.c

@@ -105,10 +105,9 @@ static int cciss_ioctl(struct inode *inode, struct file *filep,
 		unsigned int cmd, unsigned long arg);
 
 static int revalidate_allvol(kdev_t dev);
-static int revalidate_logvol(kdev_t dev, int maxusage);
 static int cciss_revalidate(kdev_t dev);
 static int deregister_disk(int ctlr, int logvol);
-static int register_new_disk(kdev_t dev, int cltr);
+static int register_new_disk(int cltr);
 
 static void cciss_getgeometry(int cntl_num);
 
@@ -333,27 +332,6 @@ static void cmd_free(ctlr_info_t *h, CommandList_struct *c, int got_from_pool)
         }
 }
 
-/*  
- * fills in the disk information. 
- */
-static void cciss_geninit( int ctlr)
-{
-	drive_info_struct *drv;
-	int i;
-	
-	/* Loop through each real device */ 
-	hba[ctlr]->gendisk.nr_real = 0; 
-	for(i=0; i< NWD; i++)
-	{
-		drv = &(hba[ctlr]->drv[i]);
-		if( !(drv->nr_blocks))
-			continue;
-		hba[ctlr]->hd[i << NWD_SHIFT].nr_sects = drv->nr_blocks;
-		//hba[ctlr]->gendisk.nr_real++;
-		(BLK_DEFAULT_QUEUE(MAJOR_NR + ctlr))->hardsect_size = drv->block_size;
-	}
-	hba[ctlr]->gendisk.nr_real = hba[ctlr]->highest_lun+1;
-}
 /*
  * Open.  Make sure the device is really there.
  */
@@ -599,7 +577,7 @@ static int cciss_ioctl(struct inode *inode, struct file *filep,
 
 	case CCISS_REGNEWD:
 	{
-		return(register_new_disk(inode->i_rdev, ctlr));
+		return(register_new_disk(ctlr));
 	}	
 	case CCISS_PASSTHRU:
 	{
@@ -726,49 +704,11 @@ static int cciss_revalidate(kdev_t dev)
 {
         int ctlr = major(dev) - MAJOR_NR;
 	int target = minor(dev) >> NWD_SHIFT;
-        struct gendisk *gdev = &(hba[ctlr]->gendisk);
-	gdev->part[minor(dev)].nr_sects = hba[ctlr]->drv[target].nr_blocks;
+        struct gendisk *disk = &hba[ctlr]->gendisk[target];
+	disk->part[0].nr_sects = hba[ctlr]->drv[target].nr_blocks;
 	return 0;
 }
 
-/* Borrowed and adapted from sd.c */
-/*
- * FIXME: we are missing the exclusion with ->open() here - it can happen
- * just as we are rereading partition tables.
- */
-static int revalidate_logvol(kdev_t dev, int maxusage)
-{
-        int ctlr, target;
-        struct gendisk *gdev;
-        unsigned long flags;
-        int res;
-
-        target = minor(dev) >> NWD_SHIFT;
-        ctlr = major(dev) - MAJOR_NR;
-        gdev = &(hba[ctlr]->gendisk);
-
-        spin_lock_irqsave(CCISS_LOCK(ctlr), flags);
-        if (hba[ctlr]->drv[target].usage_count > maxusage) {
-                spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags);
-                printk(KERN_WARNING "cciss: Device busy for "
-                        "revalidation (usage=%d)\n",
-                        hba[ctlr]->drv[target].usage_count);
-                return -EBUSY;
-        }
-        hba[ctlr]->drv[target].usage_count++;
-        spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags);
-
-	res = wipe_partitions(dev);
-	if (res)
-		goto leave;
-
-	/* setup partitions per disk */
-	grok_partitions(dev, hba[ctlr]->drv[target].nr_blocks);
-leave:
-        hba[ctlr]->drv[target].usage_count--;
-        return res;
-}
-
 /*
  * revalidate_allvol is for online array config utilities.  After a
  * utility reconfigures the drives in the array, it can use this function
@@ -799,6 +739,15 @@ static int revalidate_allvol(kdev_t dev)
         hba[ctlr]->usage_count++;
 	spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags);
 
+	for(i=0; i< NWD; i++) {
+		struct gendisk *disk = &hba[ctlr]->gendisk[i];
+		if (disk->major_name) {
+			wipe_partitions(mk_kdev(disk->major, disk->first_minor));
+			del_gendisk(disk);
+			disk->major_name = NULL;
+		}
+	}
+
         /*
          * Set the partition and block size structures for all volumes
          * on this controller to zero.  We will reread all of this data
@@ -806,8 +755,6 @@ static int revalidate_allvol(kdev_t dev)
 	memset(hba[ctlr]->hd,         0, sizeof(struct hd_struct) * 256);
         memset(hba[ctlr]->drv,        0, sizeof(drive_info_struct)
 						* CISS_MAX_LUN);
-        hba[ctlr]->gendisk.nr_real = 0;
-
         /*
          * Tell the array controller not to give us any interrupts while
          * we check the new geometry.  Then turn interrupts back on when
@@ -817,13 +764,20 @@ static int revalidate_allvol(kdev_t dev)
         cciss_getgeometry(ctlr);
         hba[ctlr]->access.set_intr_mask(hba[ctlr], CCISS_INTR_ON);
 
-        cciss_geninit(ctlr);
-        for(i=0; i<NWD; i++) {
-		kdev_t kdev = mk_kdev(major(dev), i << NWD_SHIFT);
-                if (hba[ctlr]->gendisk.part[ i<<NWD_SHIFT ].nr_sects)
-                        revalidate_logvol(kdev, 2);
+	/* Loop through each real device */ 
+	for (i = 0; i < NWD; i++) {
+		struct gendisk *disk = &hba[ctlr]->gendisk[i];
+		drive_info_struct *drv = &(hba[ctlr]->drv[i]);
+		if (!drv->nr_blocks)
+			continue;
+		(BLK_DEFAULT_QUEUE(MAJOR_NR + ctlr))->hardsect_size = drv->block_size;
+		disk->major_name = hba[ctlr]->names + 12 * i;
+		add_gendisk(disk);
+		register_disk(disk,
+			      mk_kdev(disk->major, disk->first_minor),
+			      1<<disk->minor_shift, disk->fops,
+			      drv->nr_blocks);
 	}
-
         hba[ctlr]->usage_count--;
         return 0;
 }
@@ -831,18 +785,15 @@ static int revalidate_allvol(kdev_t dev)
 static int deregister_disk(int ctlr, int logvol)
 {
 	unsigned long flags;
-	struct gendisk *gdev = &(hba[ctlr]->gendisk);
+	struct gendisk *disk = &hba[ctlr]->gendisk[logvol];
 	ctlr_info_t  *h = hba[ctlr];
-	int start, max_p;
-
 
 	if (!capable(CAP_SYS_RAWIO))
 		return -EPERM;
 
 	spin_lock_irqsave(CCISS_LOCK(ctlr), flags);
 	/* make sure logical volume is NOT is use */
-	if( h->drv[logvol].usage_count > 1)
-	{
+	if( h->drv[logvol].usage_count > 1) {
 		spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags);
                 return -EBUSY;
 	}
@@ -850,25 +801,24 @@ static int deregister_disk(int ctlr, int logvol)
 	spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags);
 
 	/* invalidate the devices and deregister the disk */ 
-	max_p = 1 << gdev->minor_shift;
-	start = logvol << gdev->minor_shift;
-	wipe_partitions(mk_kdev(MAJOR_NR+ctlr, start));
+	if (disk->major_name) {
+		wipe_partitions(mk_kdev(disk->major, disk->first_minor));
+		del_gendisk(disk);
+		disk->major_name = NULL;
+	}
 	/* check to see if it was the last disk */
-	if (logvol == h->highest_lun)
-	{
+	if (logvol == h->highest_lun) {
 		/* if so, find the new hightest lun */
 		int i, newhighest =-1;
-		for(i=0; i<h->highest_lun; i++)
-		{
+		for(i=0; i<h->highest_lun; i++) {
 			/* if the disk has size > 0, it is available */
-			if (h->gendisk.part[i << gdev->minor_shift].nr_sects)
+			if (h->drv[i].nr_blocks)
 				newhighest = i;
 		}
 		h->highest_lun = newhighest;
 				
 	}
 	--h->num_luns;
-	gdev->nr_real = h->highest_lun+1; 
 	/* zero out the disk size info */ 
 	h->drv[logvol].nr_blocks = 0;
 	h->drv[logvol].block_size = 0;
@@ -1065,11 +1015,11 @@ case CMD_HARDWARE_ERR:
         return(return_status);
 
 }
-static int register_new_disk(kdev_t dev, int ctlr)
+static int register_new_disk(int ctlr)
 {
-        struct gendisk *gdev = &(hba[ctlr]->gendisk);
+        struct gendisk *disk;
         ctlr_info_t  *h = hba[ctlr];
-        int start, max_p, i;
+        int i;
 	int num_luns;
 	int logvol;
 	int new_lun_found = 0;
@@ -1084,7 +1034,6 @@ static int register_new_disk(kdev_t dev, int ctlr)
 	__u32 lunid = 0;
 	unsigned int block_size;
 	unsigned int total_size;
-	kdev_t kdev;
 
         if (!capable(CAP_SYS_RAWIO))
                 return -EPERM;
@@ -1292,15 +1241,16 @@ static int register_new_disk(kdev_t dev, int ctlr)
 			hba[ctlr]->drv[logvol].sectors,
 			hba[ctlr]->drv[logvol].cylinders);
 	hba[ctlr]->drv[logvol].usage_count = 0;
-	max_p = 1 << gdev->minor_shift;
-        start = logvol<< gdev->minor_shift;
-	kdev = mk_kdev(MAJOR_NR + ctlr, logvol<< gdev->minor_shift);
-
-	wipe_partitions(kdev);
 	++hba[ctlr]->num_luns;
-	gdev->nr_real = hba[ctlr]->highest_lun + 1;
 	/* setup partitions per disk */
-	grok_partitions(kdev, hba[ctlr]->drv[logvol].nr_blocks);
+        disk = &hba[ctlr]->gendisk[logvol];
+	disk->major_name = hba[ctlr]->names + 12 * logvol;
+	add_gendisk(disk);
+	register_disk(disk,
+		      mk_kdev(disk->major, disk->first_minor),
+		      1<<disk->minor_shift,
+		      disk->fops,
+		      hba[ctlr]->drv[logvol].nr_blocks);
 	
 	kfree(ld_buff);
 	kfree(size_buff);
@@ -2488,22 +2438,28 @@ static int __init cciss_init_one(struct pci_dev *pdev,
 
 	blk_queue_max_sectors(q, 512);
 
-	/* Fill in the gendisk data */ 	
-	hba[i]->gendisk.major = MAJOR_NR + i;
-	hba[i]->gendisk.major_name = "cciss";
-	hba[i]->gendisk.minor_shift = NWD_SHIFT;
-	hba[i]->gendisk.part = hba[i]->hd;
-	hba[i]->gendisk.nr_real = hba[i]->highest_lun+1;  
 
-	/* Get on the disk list */ 
-	add_gendisk(&(hba[i]->gendisk));
+	for(j=0; j<NWD; j++) {
+		drive_info_struct *drv = &(hba[i]->drv[j]);
+		struct gendisk *disk = hba[i]->gendisk + j;
 
-	cciss_geninit(i);
-	for(j=0; j<NWD; j++)
-		register_disk(&(hba[i]->gendisk),
-			mk_kdev(MAJOR_NR+i, j <<4), 
-			MAX_PART, &cciss_fops, 
-			hba[i]->drv[j].nr_blocks);
+		sprintf(hba[i]->names + 12 * j, "cciss/c%dd%d", i, j);
+		disk->major = MAJOR_NR + i;
+		disk->first_minor = j << NWD_SHIFT;
+		disk->major_name = NULL;
+		disk->minor_shift = NWD_SHIFT;
+		disk->part = hba[i]->hd + (j << NWD_SHIFT);
+		disk->nr_real = 1;
+		if( !(drv->nr_blocks))
+			continue;
+		(BLK_DEFAULT_QUEUE(MAJOR_NR + i))->hardsect_size = drv->block_size;
+		disk->major_name = hba[i]->names + 12 * j;
+		add_gendisk(disk);
+		register_disk(disk,
+			      mk_kdev(disk->major, disk->first_minor),
+			      1<<disk->minor_shift, disk->fops,
+			      drv->nr_blocks);
+	}
 
 	cciss_register_scsi(i, 1);  /* hook ourself into SCSI subsystem */
 
@@ -2513,7 +2469,7 @@ static int __init cciss_init_one(struct pci_dev *pdev,
 static void __devexit cciss_remove_one (struct pci_dev *pdev)
 {
 	ctlr_info_t *tmp_ptr;
-	int i;
+	int i, j;
 	char flush_buf[4];
 	int return_code; 
 
@@ -2548,7 +2504,13 @@ static void __devexit cciss_remove_one (struct pci_dev *pdev)
 	remove_proc_entry(hba[i]->devname, proc_cciss);	
 	
 	/* remove it from the disk list */
-	del_gendisk(&(hba[i]->gendisk));
+	for (j = 0; j < NWD; j++) {
+		struct gendisk *disk = &hba[i]->gendisk[j];
+		if (disk->major_name) {
+			del_gendisk(disk);
+			disk->major_name = NULL;
+		}
+	}
 
 	pci_free_consistent(hba[i]->pdev, NR_CMDS * sizeof(CommandList_struct),
 			    hba[i]->cmd_pool, hba[i]->cmd_pool_dhandle);

drivers/block/cciss.h

@@ -81,7 +81,8 @@ struct ctlr_info
 	int			nr_frees; 
 
 	// Disk structures we need to pass back
-	struct gendisk   gendisk;
+	struct gendisk   gendisk[NWD];
+	char names[12 * NWD];
 	   // indexed by minor numbers
 	struct hd_struct hd[256];
 	int		 sizes[256];

drivers/block/cpqarray.c

@@ -1483,6 +1483,7 @@ static int revalidate_allvol(kdev_t dev)
 		if (!drv->nr_blks)
 			continue;
 		(BLK_DEFAULT_QUEUE(MAJOR_NR + ctlr))->hardsect_size = drv->blk_size;
+		disk->major_name = ida_names + (ctlr*NWD+i)*10;
 		add_gendisk(disk);
 		register_disk(disk,
 			      mk_kdev(disk->major,disk->first_minor),

drivers/char/Config.help

@@ -969,6 +969,31 @@ CONFIG_RTC
   The module is called rtc.o. If you want to compile it as a module,
   say M here and read <file:Documentation/modules.txt>.
 
+Generic Real Time Clock Support
+CONFIG_GEN_RTC
+  If you say Y here and create a character special file /dev/rtc with
+  major number 10 and minor number 135 using mknod ("man mknod"), you
+  will get access to the real time clock (or hardware clock) built
+  into your computer.
+
+  It reports status information via the file /proc/driver/rtc and its
+  behaviour is set by various ioctls on /dev/rtc. If you enable the
+  "extended RTC operation" below it will also provide an emulation
+  for RTC_UIE which is required by some programs and may improve
+  precision in some cases.
+
+  This driver is also available as a module ( = code which can be
+  inserted in and removed from the running kernel whenever you want).
+  The module is called genrtc.o. If you want to compile it as a module,
+  say M here and read <file:Documentation/modules.txt>. To load the
+  module automaticaly add 'alias char-major-10-135 genrtc' to your
+  /etc/modules.conf
+
+Extended RTC operation
+CONFIG_GEN_RTC_X
+  Provides an emulation for RTC_UIE which is required by some programs
+  and may improve precision of the generic RTC support in some cases.
+
 CONFIG_H8
   The Hitachi H8/337 is a microcontroller used to deal with the power
   and thermal environment. If you say Y here, you will be able to

drivers/char/Config.in

@@ -151,6 +151,12 @@ if [ "$CONFIG_X86" = "y" -o "$CONFIG_IA64" = "y" ]; then
 fi
 tristate '/dev/nvram support' CONFIG_NVRAM
 tristate 'Enhanced Real Time Clock Support' CONFIG_RTC
+if [ "$CONFIG_RTC" != "y" ]; then
+   tristate 'Generic /dev/rtc emulation' CONFIG_GEN_RTC      
+   if [ "$CONFIG_GEN_RTC" != "n" ]; then
+      bool '   Extended RTC operation' CONFIG_GEN_RTC_X
+   fi
+fi
 if [ "$CONFIG_IA64" = "y" ]; then
    bool 'EFI Real Time Clock Services' CONFIG_EFI_RTC
 fi

drivers/char/Makefile

@@ -167,6 +167,7 @@ obj-$(CONFIG_APPLICOM) += applicom.o
 obj-$(CONFIG_SONYPI) += sonypi.o
 obj-$(CONFIG_ATARIMOUSE) += atarimouse.o
 obj-$(CONFIG_RTC) += rtc.o
+obj-$(CONFIG_GEN_RTC) += genrtc.o
 obj-$(CONFIG_EFI_RTC) += efirtc.o
 ifeq ($(CONFIG_PPC),)
   obj-$(CONFIG_NVRAM) += nvram.o

drivers/char/genrtc.c

+/*
+ *	Real Time Clock interface for q40 and other m68k machines
+ *      emulate some RTC irq capabilities in software
+ *
+ *      Copyright (C) 1999 Richard Zidlicky
+ *
+ *	based on Paul Gortmaker's rtc.c device and
+ *           Sam Creasey Generic rtc driver
+ *
+ *	This driver allows use of the real time clock (built into
+ *	nearly all computers) from user space. It exports the /dev/rtc
+ *	interface supporting various ioctl() and also the /proc/dev/rtc
+ *	pseudo-file for status information.
+ *
+ *	The ioctls can be used to set the interrupt behaviour where
+ *  supported.
+ *
+ *	The /dev/rtc interface will block on reads until an interrupt
+ *	has been received. If a RTC interrupt has already happened,
+ *	it will output an unsigned long and then block. The output value
+ *	contains the interrupt status in the low byte and the number of
+ *	interrupts since the last read in the remaining high bytes. The
+ *	/dev/rtc interface can also be used with the select(2) call.
+ *
+ *	This program is free software; you can redistribute it and/or
+ *	modify it under the terms of the GNU General Public License
+ *	as published by the Free Software Foundation; either version
+ *	2 of the License, or (at your option) any later version.
+ *
+
+ *      1.01 fix for 2.3.X                    rz@linux-m68k.org
+ *      1.02 merged with code from genrtc.c   rz@linux-m68k.org
+ *      1.03 make it more portable            zippel@linux-m68k.org
+ *      1.04 removed useless timer code       rz@linux-m68k.org
+ *      1.05 portable RTC_UIE emulation       rz@linux-m68k.org
+ *      1.06 set_rtc_time can return an error trini@kernel.crashing.org
+ */
+
+#define RTC_VERSION	"1.06"
+
+#include <linux/module.h>
+#include <linux/config.h>
+#include <linux/errno.h>
+#include <linux/miscdevice.h>
+#include <linux/fcntl.h>
+
+#include <linux/rtc.h>
+#include <linux/init.h>
+#include <linux/poll.h>
+#include <linux/proc_fs.h>
+#include <linux/tqueue.h>
+
+#include <asm/uaccess.h>
+#include <asm/system.h>
+#include <asm/rtc.h>
+
+/*
+ *	We sponge a minor off of the misc major. No need slurping
+ *	up another valuable major dev number for this. If you add
+ *	an ioctl, make sure you don't conflict with SPARC's RTC
+ *	ioctls.
+ */
+
+static DECLARE_WAIT_QUEUE_HEAD(gen_rtc_wait);
+
+static int gen_rtc_ioctl(struct inode *inode, struct file *file,
+		     unsigned int cmd, unsigned long arg);
+
+/*
+ *	Bits in gen_rtc_status.
+ */
+
+#define RTC_IS_OPEN		0x01	/* means /dev/rtc is in use	*/
+
+unsigned char gen_rtc_status;		/* bitmapped status byte.	*/
+unsigned long gen_rtc_irq_data;		/* our output to the world	*/
+
+/* months start at 0 now */
+unsigned char days_in_mo[] =
+{31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31};
+
+static int irq_active;
+
+#ifdef CONFIG_GEN_RTC_X
+struct tq_struct genrtc_task;
+static struct timer_list timer_task;
+
+static unsigned int oldsecs;
+static int lostint;
+static int tt_exp;
+
+void gen_rtc_timer(unsigned long data);
+
+static volatile int stask_active;              /* schedule_task */
+static volatile int ttask_active;              /* timer_task */
+static int stop_rtc_timers;                    /* don't requeue tasks */
+static spinlock_t gen_rtc_lock = SPIN_LOCK_UNLOCKED;
+
+/*
+ * Routine to poll RTC seconds field for change as often as posible,
+ * after first RTC_UIE use timer to reduce polling
+ */
+void genrtc_troutine(void *data)
+{
+	unsigned int tmp = get_rtc_ss();
+	
+	if (stop_rtc_timers) {
+		stask_active = 0;
+		return;
+	}
+
+	if (oldsecs != tmp){
+		oldsecs = tmp;
+
+		timer_task.function = gen_rtc_timer;
+		timer_task.expires = jiffies + HZ - (HZ/10);
+		tt_exp=timer_task.expires;
+		ttask_active=1;
+		stask_active=0;
+		add_timer(&timer_task);
+
+		gen_rtc_interrupt(0);
+	} else if (schedule_task(&genrtc_task) == 0)
+		stask_active = 0;
+}
+
+void gen_rtc_timer(unsigned long data)
+{
+	lostint = get_rtc_ss() - oldsecs ;
+	if (lostint<0) 
+		lostint = 60 - lostint;
+	if (time_after(jiffies, tt_exp))
+		printk(KERN_INFO "genrtc: timer task delayed by %ld jiffies\n",
+		       jiffies-tt_exp);
+	ttask_active=0;
+	stask_active=1;
+	if ((schedule_task(&genrtc_task) == 0))
+		stask_active = 0;
+}
+
+/* 
+ * call gen_rtc_interrupt function to signal an RTC_UIE,
+ * arg is unused.
+ * Could be invoked either from a real interrupt handler or
+ * from some routine that periodically (eg 100HZ) monitors
+ * whether RTC_SECS changed
+ */
+void gen_rtc_interrupt(unsigned long arg)
+{
+	/*  We store the status in the low byte and the number of
+	 *	interrupts received since the last read in the remainder
+	 *	of rtc_irq_data.  */
+
+	gen_rtc_irq_data += 0x100;
+	gen_rtc_irq_data &= ~0xff;
+	gen_rtc_irq_data |= RTC_UIE;
+
+	if (lostint){
+		printk("genrtc: system delaying clock ticks?\n");
+		/* increment count so that userspace knows something is wrong */
+		gen_rtc_irq_data += ((lostint-1)<<8);
+		lostint = 0;
+	}
+
+	wake_up_interruptible(&gen_rtc_wait);
+}
+
+/*
+ *	Now all the various file operations that we export.
+ */
+static ssize_t gen_rtc_read(struct file *file, char *buf,
+			size_t count, loff_t *ppos)
+{
+	DECLARE_WAITQUEUE(wait, current);
+	unsigned long data;
+	ssize_t retval;
+
+        if (count != sizeof (unsigned int) && count != sizeof (unsigned long))
+		return -EINVAL;
+
+	if (file->f_flags & O_NONBLOCK && !gen_rtc_irq_data)
+		return -EAGAIN;
+
+	add_wait_queue(&gen_rtc_wait, &wait);
+	retval = -ERESTARTSYS;
+
+	while (1) {
+		set_current_state(TASK_INTERRUPTIBLE);
+		data = xchg(&gen_rtc_irq_data, 0);
+		if (data)
+			break;
+		if (signal_pending(current))
+			goto out;
+		schedule();
+	}
+
+	/* first test allows optimizer to nuke this case for 32-bit machines */
+	if (sizeof (int) != sizeof (long) && count == sizeof (unsigned int)) {
+		unsigned int uidata = data;
+		retval = put_user(uidata, (unsigned long *)buf);
+	}
+	else {
+		retval = put_user(data, (unsigned long *)buf);
+	}
+	if (!retval)
+		retval = sizeof(unsigned long);
+ out:
+	current->state = TASK_RUNNING;
+	remove_wait_queue(&gen_rtc_wait, &wait);
+
+	return retval;
+}
+
+static unsigned int gen_rtc_poll(struct file *file,
+				 struct poll_table_struct *wait)
+{
+	poll_wait(file, &gen_rtc_wait, wait);
+	if (gen_rtc_irq_data != 0)
+		return POLLIN | POLLRDNORM;
+	return 0;
+}
+
+#endif
+
+/*
+ * Used to disable/enable interrupts, only RTC_UIE supported
+ * We also clear out any old irq data after an ioctl() that
+ * meddles with the interrupt enable/disable bits.
+ */
+
+static inline void gen_clear_rtc_irq_bit(unsigned char bit)
+{
+#ifdef CONFIG_GEN_RTC_X
+	stop_rtc_timers = 1;
+	if (ttask_active){
+		del_timer_sync(&timer_task);
+		ttask_active = 0;
+	}
+	while (stask_active)
+		schedule();
+
+	spin_lock(&gen_rtc_lock);
+	irq_active = 0;
+	spin_unlock(&gen_rtc_lock);
+#endif
+}
+
+static inline int gen_set_rtc_irq_bit(unsigned char bit)
+{
+#ifdef CONFIG_GEN_RTC_X
+	spin_lock(&gen_rtc_lock);
+	if ( !irq_active ) {
+		irq_active = 1;
+		stop_rtc_timers = 0;
+		lostint = 0;
+		genrtc_task.routine = genrtc_troutine;
+		oldsecs = get_rtc_ss();
+		init_timer(&timer_task);
+
+		stask_active = 1;
+		if (schedule_task(&genrtc_task) == 0){
+			stask_active = 0;
+		}
+	}
+	spin_unlock(&gen_rtc_lock);
+	gen_rtc_irq_data = 0;
+	return 0;
+#else
+	return -EINVAL;
+#endif
+}
+
+static int gen_rtc_ioctl(struct inode *inode, struct file *file,
+			 unsigned int cmd, unsigned long arg)
+{
+	struct rtc_time wtime;
+	struct rtc_pll_info pll;
+
+	switch (cmd) {
+
+	case RTC_PLL_GET:
+	    if (get_rtc_pll(&pll))
+	 	    return -EINVAL;
+	    else
+		    return copy_to_user((void *)arg, &pll, sizeof pll) ? -EFAULT : 0;
+
+	case RTC_PLL_SET:
+		if (!capable(CAP_SYS_TIME))
+			return -EACCES;
+		if (copy_from_user(&pll, (struct rtc_pll_info*)arg,
+				   sizeof(pll)))
+			return -EFAULT;
+	    return set_rtc_pll(&pll);
+
+	case RTC_UIE_OFF:	/* disable ints from RTC updates.	*/
+		gen_clear_rtc_irq_bit(RTC_UIE);
+		return 0;
+
+	case RTC_UIE_ON:	/* enable ints for RTC updates.	*/
+	        return gen_set_rtc_irq_bit(RTC_UIE);
+
+	case RTC_RD_TIME:	/* Read the time/date from RTC	*/
+		/* this doesn't get week-day, who cares */
+		memset(&wtime, 0, sizeof(wtime));
+		get_rtc_time(&wtime);
+
+		return copy_to_user((void *)arg, &wtime, sizeof(wtime)) ? -EFAULT : 0;
+
+	case RTC_SET_TIME:	/* Set the RTC */
+	    {
+		int year;
+		unsigned char leap_yr;
+
+		if (!capable(CAP_SYS_TIME))
+			return -EACCES;
+
+		if (copy_from_user(&wtime, (struct rtc_time *)arg,
+				   sizeof(wtime)))
+			return -EFAULT;
+
+		year = wtime.tm_year + 1900;
+		leap_yr = ((!(year % 4) && (year % 100)) ||
+			   !(year % 400));
+
+		if ((wtime.tm_mon < 0 || wtime.tm_mon > 11) || (wtime.tm_mday < 1))
+			return -EINVAL;
+
+		if (wtime.tm_mday < 0 || wtime.tm_mday >
+		    (days_in_mo[wtime.tm_mon] + ((wtime.tm_mon == 1) && leap_yr)))
+			return -EINVAL;
+
+		if (wtime.tm_hour < 0 || wtime.tm_hour >= 24 ||
+		    wtime.tm_min < 0 || wtime.tm_min >= 60 ||
+		    wtime.tm_sec < 0 || wtime.tm_sec >= 60)
+			return -EINVAL;
+
+		return set_rtc_time(&wtime);
+	    }
+	}
+
+	return -EINVAL;
+}
+
+/*
+ *	We enforce only one user at a time here with the open/close.
+ *	Also clear the previous interrupt data on an open, and clean
+ *	up things on a close.
+ */
+
+static int gen_rtc_open(struct inode *inode, struct file *file)
+{
+	if (gen_rtc_status & RTC_IS_OPEN)
+		return -EBUSY;
+
+	MOD_INC_USE_COUNT;
+
+	gen_rtc_status |= RTC_IS_OPEN;
+	gen_rtc_irq_data = 0;
+	irq_active = 0;
+
+	return 0;
+}
+
+static int gen_rtc_release(struct inode *inode, struct file *file)
+{
+	/*
+	 * Turn off all interrupts once the device is no longer
+	 * in use and clear the data.
+	 */
+
+	gen_clear_rtc_irq_bit(RTC_PIE|RTC_AIE|RTC_UIE);
+
+	gen_rtc_status &= ~RTC_IS_OPEN;
+	MOD_DEC_USE_COUNT;
+
+	return 0;
+}
+
+static int gen_rtc_read_proc(char *page, char **start, off_t off,
+			     int count, int *eof, void *data);
+
+
+/*
+ *	The various file operations we support.
+ */
+
+static struct file_operations gen_rtc_fops = {
+	.owner =	THIS_MODULE,
+#ifdef CONFIG_GEN_RTC_X
+	.read =		gen_rtc_read,
+	.poll =		gen_rtc_poll,
+#endif
+	.ioctl =	gen_rtc_ioctl,
+	.open =		gen_rtc_open,
+	.release =	gen_rtc_release
+};
+
+static struct miscdevice rtc_gen_dev =
+{
+	RTC_MINOR,
+	"rtc",
+	&gen_rtc_fops
+};
+
+int __init rtc_generic_init(void)
+{
+
+		printk(KERN_INFO "Generic RTC Driver v%s\n", RTC_VERSION);
+
+	misc_register(&rtc_gen_dev);
+	create_proc_read_entry ("driver/rtc", 0, 0, gen_rtc_read_proc, NULL);
+
+	return 0;
+}
+
+static void __exit rtc_generic_exit(void)
+{
+	remove_proc_entry ("driver/rtc", NULL);
+	misc_deregister(&rtc_gen_dev);
+}
+
+module_init(rtc_generic_init);
+module_exit(rtc_generic_exit);
+EXPORT_NO_SYMBOLS;
+
+
+/*
+ *	Info exported via "/proc/rtc".
+ */
+
+int gen_rtc_proc_output(char *buf)
+{
+	char *p;
+	struct rtc_time tm;
+	unsigned tmp;
+	struct rtc_pll_info pll;
+
+	p = buf;
+
+	get_rtc_time(&tm);
+
+	p += sprintf(p,
+		     "rtc_time\t: %02d:%02d:%02d\n"
+		     "rtc_date\t: %04d-%02d-%02d\n"
+		     "rtc_epoch\t: %04u\n",
+		     tm.tm_hour, tm.tm_min, tm.tm_sec,
+		     tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday, 1900);
+
+	tm.tm_hour=0;tm.tm_min=0;tm.tm_sec=0;
+
+	p += sprintf(p, "alarm\t\t: ");
+	if (tm.tm_hour <= 24)
+		p += sprintf(p, "%02d:", tm.tm_hour);
+	else
+		p += sprintf(p, "**:");
+
+	if (tm.tm_min <= 59)
+		p += sprintf(p, "%02d:", tm.tm_min);
+	else
+		p += sprintf(p, "**:");
+
+	if (tm.tm_sec <= 59)
+		p += sprintf(p, "%02d\n", tm.tm_sec);
+	else
+		p += sprintf(p, "**\n");
+
+	tmp= RTC_24H ;
+	p += sprintf(p,
+		     "DST_enable\t: %s\n"
+		     "BCD\t\t: %s\n"
+		     "24hr\t\t: %s\n"
+		     "square_wave\t: %s\n"
+		     "alarm_IRQ\t: %s\n"
+		     "update_IRQ\t: %s\n"
+		     "periodic_IRQ\t: %s\n"
+		     "periodic_freq\t: %ld\n"
+		     "batt_status\t: %s\n",
+		     (tmp & RTC_DST_EN) ? "yes" : "no",
+		     (tmp & RTC_DM_BINARY) ? "no" : "yes",
+		     (tmp & RTC_24H) ? "yes" : "no",
+		     (tmp & RTC_SQWE) ? "yes" : "no",
+		     (tmp & RTC_AIE) ? "yes" : "no",
+		     irq_active ? "yes" : "no",
+		     (tmp & RTC_PIE) ? "yes" : "no",
+		     0L /* freq */,
+		     "okay" );
+	if (!get_rtc_pll(&pll))
+	    p += sprintf(p,
+			 "PLL adjustment\t: %d\n"
+			 "PLL max +ve adjustment\t: %d\n"
+			 "PLL max -ve adjustment\t: %d\n"
+			 "PLL +ve adjustment factor\t: %d\n"
+			 "PLL -ve adjustment factor\t: %d\n"
+			 "PLL frequency\t: %ld\n",
+			 pll.pll_value,
+			 pll.pll_max,
+			 pll.pll_min,
+			 pll.pll_posmult,
+			 pll.pll_negmult,
+			 pll.pll_clock);
+	return  p - buf;
+}
+
+static int gen_rtc_read_proc(char *page, char **start, off_t off,
+			     int count, int *eof, void *data)
+{
+	int len = gen_rtc_proc_output (page);
+        if (len <= off+count) *eof = 1;
+	*start = page + off;
+	len -= off;
+        if (len>count) len = count;
+        if (len<0) len = 0;
+	return len;
+}
+
+
+MODULE_AUTHOR("Richard Zidlicky");
+MODULE_LICENSE("GPL");

drivers/char/mem.c

@@ -210,6 +210,9 @@ static int mmap_mem(struct file * file, struct vm_area_struct * vma)
 	return 0;
 }
 
+extern long vread(char *buf, char *addr, unsigned long count);
+extern long vwrite(char *buf, char *addr, unsigned long count);
+
 /*
  * This function reads the *virtual* memory as seen by the kernel.
  */
@@ -273,8 +276,6 @@ static ssize_t read_kmem(struct file *file, char *buf,
  	return virtr + read;
 }
 
-extern long vwrite(char *buf, char *addr, unsigned long count);
-
 /*
  * This function writes to the *virtual* memory as seen by the kernel.
  */

drivers/i2c/Config.in

@@ -32,10 +32,10 @@ if [ "$CONFIG_I2C" != "n" ]; then
          dep_tristate '  Embedded Planet RPX Lite/Classic suppoort' CONFIG_I2C_RPXLITE $CONFIG_I2C_ALGO8XX
       fi
    fi
-   if [ "$CONFIG_405" = "y" ]; then
-      dep_tristate 'PPC 405 I2C Algorithm' CONFIG_I2C_PPC405_ALGO $CONFIG_I2C
-      if [ "$CONFIG_I2C_PPC405_ALGO" != "n" ]; then
-         dep_tristate '  PPC 405 I2C Adapter' CONFIG_I2C_PPC405_ADAP $CONFIG_I2C_PPC405_ALGO
+   if [ "$CONFIG_IBM_OCP" = "y" ]; then
+      dep_tristate 'IBM on-chip I2C Algorithm' CONFIG_I2C_IBM_OCP_ALGO $CONFIG_I2C
+      if [ "$CONFIG_I2C_IBM_OCP_ALGO" != "n" ]; then
+         dep_tristate '  IBM on-chip I2C Adapter' CONFIG_I2C_IBM_OCP_ADAP $CONFIG_I2C_IBM_OCP_ALGO
       fi
    fi
 

drivers/scsi/sym53c8xx_2/sym_glue.c

@@ -1107,7 +1107,6 @@ static void sym_eh_timeout(u_long p) { __sym_eh_done((Scsi_Cmnd *)p, 1); }
 static int sym_eh_handler(int op, char *opname, Scsi_Cmnd *cmd)
 {
 	hcb_p np = SYM_SOFTC_PTR(cmd);
-	unsigned long flags;
 	SYM_QUEHEAD *qp;
 	int to_do = SYM_EH_DO_IGNORE;
 	int sts = -1;
@@ -1118,8 +1117,6 @@ static int sym_eh_handler(int op, char *opname, Scsi_Cmnd *cmd)
 
 	printf_warning("%s: %s operation started.\n", devname, opname);
 
-	SYM_LOCK_HCB(np, flags);
-
 #if 0
 	/* This one should be the result of some race, thus to ignore */
 	if (cmd->serial_number != cmd->serial_number_at_timeout)
@@ -1198,8 +1195,6 @@ static int sym_eh_handler(int op, char *opname, Scsi_Cmnd *cmd)
 	if (to_do == SYM_EH_DO_COMPLETE)
 		sym_xpt_done2(np, cmd, CAM_REQ_ABORTED);
 
-	SYM_UNLOCK_HCB(np, flags);
-
 	/* Wait for completion with locks released, as required by kernel */
 	if (to_do == SYM_EH_DO_WAIT) {
 		init_timer(&ep->timer);

drivers/usb/core/Makefile

@@ -2,10 +2,10 @@
 # Makefile for USB Core files and filesystem
 #
 
-export-objs	:= usb.o hcd.o hcd-pci.o urb.o message.o file.o
+export-objs	:= usb.o hcd.o hcd-pci.o urb.o message.o file.o buffer.o
 
 usbcore-objs	:= usb.o usb-debug.o hub.o hcd.o urb.o message.o \
-			config.o file.o
+			config.o file.o buffer.o
 
 ifeq ($(CONFIG_PCI),y)
 	usbcore-objs	+= hcd-pci.o

drivers/usb/core/buffer.c

+/*
+ * DMA memory management for framework level HCD code (hc_driver)
+ *
+ * This implementation plugs in through generic "usb_bus" level methods,
+ * and works with real PCI, or when "pci device == null" makes sense.
+ */
+
+#include <linux/config.h>
+#include <linux/module.h>
+#include <linux/version.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/pci.h>
+
+
+#ifdef CONFIG_USB_DEBUG
+	#define DEBUG
+#else
+	#undef DEBUG
+#endif
+
+#include <linux/usb.h>
+#include "hcd.h"
+
+
+/*
+ * DMA-Consistent Buffers
+ */
+
+/* FIXME tune these based on pool statistics ... */
+static const size_t	pool_max [HCD_BUFFER_POOLS] = {
+	32,
+	128,
+	512,
+	PAGE_SIZE / 2
+	/* bigger --> allocate pages */
+};
+
+
+/* SETUP primitives */
+
+/**
+ * hcd_buffer_create - initialize buffer pools
+ * @hcd: the bus whose buffer pools are to be initialized
+ *
+ * Call this as part of initializing a host controller that uses the pci dma
+ * memory allocators.  It initializes some pools of dma-consistent memory that
+ * will be shared by all drivers using that controller, or returns a negative
+ * errno value on error.
+ *
+ * Call hcd_buffer_destroy() to clean up after using those pools.
+ */
+int hcd_buffer_create (struct usb_hcd *hcd)
+{
+	char		name [16];
+	int 		i, size;
+
+	for (i = 0; i < HCD_BUFFER_POOLS; i++) { 
+		if (!(size = pool_max [i]))
+			continue;
+		snprintf (name, sizeof name, "buffer-%d", size);
+		hcd->pool [i] = pci_pool_create (name, hcd->pdev,
+				size, size, 0, SLAB_KERNEL);
+		if (!hcd->pool [i]) {
+			hcd_buffer_destroy (hcd);
+			return -ENOMEM;
+		}
+	}
+	return 0;
+}
+EXPORT_SYMBOL (hcd_buffer_create);
+
+
+/**
+ * hcd_buffer_destroy - deallocate buffer pools
+ * @hcd: the bus whose buffer pools are to be destroyed
+ *
+ * This frees the buffer pools created by hcd_buffer_create().
+ */
+void hcd_buffer_destroy (struct usb_hcd *hcd)
+{
+	int		i;
+
+	for (i = 0; i < HCD_BUFFER_POOLS; i++) { 
+		struct pci_pool		*pool = hcd->pool [i];
+		if (pool) {
+			pci_pool_destroy (pool);
+			hcd->pool [i] = 0;
+		}
+	}
+}
+EXPORT_SYMBOL (hcd_buffer_destroy);
+
+
+/* sometimes alloc/free could use kmalloc with SLAB_DMA, for
+ * better sharing and to leverage mm/slab.c intelligence.
+ */
+
+void *hcd_buffer_alloc (
+	struct usb_bus 		*bus,
+	size_t			size,
+	int			mem_flags,
+	dma_addr_t		*dma
+)
+{
+	struct usb_hcd		*hcd = bus->hcpriv;
+	int 			i;
+
+	for (i = 0; i < HCD_BUFFER_POOLS; i++) {
+		if (size <= pool_max [i])
+			return pci_pool_alloc (hcd->pool [i], mem_flags, dma);
+	}
+	return pci_alloc_consistent (hcd->pdev, size, dma);
+}
+
+void hcd_buffer_free (
+	struct usb_bus 		*bus,
+	size_t			size,
+	void 			*addr,
+	dma_addr_t		dma
+)
+{
+	struct usb_hcd		*hcd = bus->hcpriv;
+	int 			i;
+
+	for (i = 0; i < HCD_BUFFER_POOLS; i++) {
+		if (size <= pool_max [i]) {
+			pci_pool_free (hcd->pool [i], addr, dma);
+			return;
+		}
+	}
+	pci_free_consistent (hcd->pdev, size, addr, dma);
+}
+
+
+/*
+ * DMA-Mappings for arbitrary memory buffers
+ */
+
+int hcd_buffer_map (
+	struct usb_bus	*bus,
+	void		*addr,
+	dma_addr_t	*dma,
+	size_t		size,
+	int		direction
+) {
+	struct usb_hcd	*hcd = bus->hcpriv;
+
+	// FIXME pci_map_single() has no standard failure mode!
+	*dma = pci_map_single (hcd->pdev, addr, size,
+			(direction == USB_DIR_IN)
+				? PCI_DMA_FROMDEVICE
+				: PCI_DMA_TODEVICE);
+	return 0;
+}
+
+void hcd_buffer_dmasync (
+	struct usb_bus	*bus,
+	dma_addr_t	dma,
+	size_t		size,
+	int		direction
+) {
+	struct usb_hcd *hcd = bus->hcpriv;
+
+	pci_dma_sync_single (hcd->pdev, dma, size,
+			(direction == USB_DIR_IN)
+				? PCI_DMA_FROMDEVICE
+				: PCI_DMA_TODEVICE);
+}
+
+void hcd_buffer_unmap (
+	struct usb_bus	*bus,
+	dma_addr_t	dma,
+	size_t		size,
+	int		direction
+) {
+	struct usb_hcd *hcd = bus->hcpriv;
+
+	pci_unmap_single (hcd->pdev, dma, size,
+			(direction == USB_DIR_IN)
+				? PCI_DMA_FROMDEVICE
+				: PCI_DMA_TODEVICE);
+}
+
+
+// FIXME DMA-Mappings for struct scatterlist

drivers/usb/core/hcd-pci.c

@@ -130,10 +130,19 @@ int usb_hcd_pci_probe (struct pci_dev *dev, const struct pci_device_id *id)
 			return retval;
 		}
 	}
-	pci_set_drvdata(dev, hcd);
+	pci_set_drvdata (dev, hcd);
 	hcd->driver = driver;
 	hcd->description = driver->description;
 	hcd->pdev = dev;
+	hcd->self.bus_name = dev->slot_name;
+	hcd->product_desc = dev->name;
+
+	if ((retval = hcd_buffer_create (hcd)) != 0) {
+clean_3:
+		driver->hcd_free (hcd);
+		goto clean_2;
+	}
+
 	info ("%s @ %s, %s", hcd->description,  dev->slot_name, dev->name);
 
 	pci_read_config_byte (dev, PCI_LATENCY_TIMER, &latency);
@@ -154,8 +163,7 @@ int usb_hcd_pci_probe (struct pci_dev *dev, const struct pci_device_id *id)
 			!= 0) {
 		err ("request interrupt %s failed", bufp);
 		retval = -EBUSY;
-		driver->hcd_free (hcd);
-		goto clean_2;
+		goto clean_3;
 	}
 	hcd->irq = dev->irq;
 
@@ -168,8 +176,6 @@ int usb_hcd_pci_probe (struct pci_dev *dev, const struct pci_device_id *id)
 	usb_bus_init (&hcd->self);
 	hcd->self.op = &usb_hcd_operations;
 	hcd->self.hcpriv = (void *) hcd;
-	hcd->self.bus_name = dev->slot_name;
-	hcd->product_desc = dev->name;
 
 	INIT_LIST_HEAD (&hcd->dev_list);
 
@@ -216,6 +222,7 @@ void usb_hcd_pci_remove (struct pci_dev *dev)
 	usb_disconnect (&hub);
 
 	hcd->driver->stop (hcd);
+	hcd_buffer_destroy (hcd);
 	hcd->state = USB_STATE_HALT;
 
 	free_irq (hcd->irq, hcd);

drivers/usb/core/hcd.c

@@ -454,7 +454,6 @@ static int rh_status_urb (struct usb_hcd *hcd, struct urb *urb)
 	/* rh_timer protected by hcd_data_lock */
 	if (timer_pending (&hcd->rh_timer)
 			|| urb->status != -EINPROGRESS
-			|| !HCD_IS_RUNNING (hcd->state)
 			|| urb->transfer_buffer_length < len) {
 		dbg ("not queuing status urb, stat %d", urb->status);
 		return -EINVAL;
@@ -508,8 +507,12 @@ static void rh_report_status (unsigned long ptr)
 				BUG ();
 			}
 			spin_unlock_irqrestore (&hcd_data_lock, flags);
-		} else
+		} else {
 			spin_unlock_irqrestore (&urb->lock, flags);
+			spin_lock_irqsave (&hcd_data_lock, flags);
+			rh_status_urb (hcd, urb);
+			spin_unlock_irqrestore (&hcd_data_lock, flags);
+		}
 	} else {
 		/* this urb's been unlinked */
 		urb->hcpriv = 0;
@@ -1245,6 +1248,11 @@ struct usb_operations usb_hcd_operations = {
 	.submit_urb =		hcd_submit_urb,
 	.unlink_urb =		hcd_unlink_urb,
 	.deallocate =		hcd_free_dev,
+	.buffer_alloc =		hcd_buffer_alloc,
+	.buffer_free =		hcd_buffer_free,
+	.buffer_map =		hcd_buffer_map,
+	.buffer_dmasync =	hcd_buffer_dmasync,
+	.buffer_unmap =		hcd_buffer_unmap,
 };
 EXPORT_SYMBOL (usb_hcd_operations);
 

drivers/usb/core/hcd.h

@@ -58,6 +58,9 @@ struct usb_hcd {	/* usb_bus.hcpriv points to this */
 	atomic_t		resume_count;	/* multiple resumes issue */
 #endif
 
+#define HCD_BUFFER_POOLS	4
+	struct pci_pool		*pool [HCD_BUFFER_POOLS];
+
 	int			state;
 #	define	__ACTIVE		0x01
 #	define	__SLEEPY		0x02
@@ -109,6 +112,25 @@ struct usb_operations {
 	int (*get_frame_number) (struct usb_device *usb_dev);
 	int (*submit_urb) (struct urb *urb, int mem_flags);
 	int (*unlink_urb) (struct urb *urb);
+
+	/* allocate dma-consistent buffer for URB_DMA_NOMAPPING */
+	void *(*buffer_alloc)(struct usb_bus *bus, size_t size,
+			int mem_flags,
+			dma_addr_t *dma);
+	void (*buffer_free)(struct usb_bus *bus, size_t size,
+			void *addr, dma_addr_t dma);
+
+	int (*buffer_map) (struct usb_bus *bus,
+		void *addr, dma_addr_t *dma,
+		size_t size, int direction);
+	void (*buffer_dmasync) (struct usb_bus *bus,
+		dma_addr_t dma,
+		size_t size, int direction);
+	void (*buffer_unmap) (struct usb_bus *bus,
+		dma_addr_t dma,
+		size_t size, int direction);
+
+	// FIXME  also: buffer_sg_map (), buffer_sg_unmap ()
 };
 
 /* each driver provides one of these, and hardware init support */
@@ -181,6 +203,25 @@ extern int usb_hcd_pci_resume (struct pci_dev *dev);
 
 #endif /* CONFIG_PCI */
 
+/* pci-ish (pdev null is ok) buffer alloc/mapping support */
+int hcd_buffer_create (struct usb_hcd *hcd);
+void hcd_buffer_destroy (struct usb_hcd *hcd);
+
+void *hcd_buffer_alloc (struct usb_bus *bus, size_t size,
+	int mem_flags, dma_addr_t *dma);
+void hcd_buffer_free (struct usb_bus *bus, size_t size,
+	void *addr, dma_addr_t dma);
+
+int hcd_buffer_map (struct usb_bus *bus,
+	void *addr, dma_addr_t *dma,
+	size_t size, int direction);
+void hcd_buffer_dmasync (struct usb_bus *bus,
+	dma_addr_t dma,
+	size_t size, int direction);
+void hcd_buffer_unmap (struct usb_bus *bus,
+	dma_addr_t dma,
+	size_t size, int direction);
+
 /* generic bus glue, needed for host controllers that don't use PCI */
 extern struct usb_operations usb_hcd_operations;
 extern void usb_hcd_irq (int irq, void *__hcd, struct pt_regs *r);

drivers/usb/core/usb.c

@@ -863,9 +863,11 @@ static ssize_t show_product (struct device *dev, char *buf, size_t count, loff_t
 		return 0;
 	udev = to_usb_device (dev);
 
-	len = usb_string(udev, udev->descriptor.iProduct, buf, PAGE_SIZE); 
+	len = usb_string(udev, udev->descriptor.iProduct, buf, PAGE_SIZE);
+	if (len < 0)
+		return 0;
 	buf[len] = '\n';
-	buf[len+1] = 0x00;
+	buf[len+1] = 0;
 	return len+1;
 }
 static DEVICE_ATTR(product,"product",S_IRUGO,show_product,NULL);
@@ -881,9 +883,11 @@ show_manufacturer (struct device *dev, char *buf, size_t count, loff_t off)
 		return 0;
 	udev = to_usb_device (dev);
 
-	len = usb_string(udev, udev->descriptor.iManufacturer, buf, PAGE_SIZE); 
+	len = usb_string(udev, udev->descriptor.iManufacturer, buf, PAGE_SIZE);
+	if (len < 0)
+		return 0;
 	buf[len] = '\n';
-	buf[len+1] = 0x00;
+	buf[len+1] = 0;
 	return len+1;
 }
 static DEVICE_ATTR(manufacturer,"manufacturer",S_IRUGO,show_manufacturer,NULL);
@@ -899,9 +903,11 @@ show_serial (struct device *dev, char *buf, size_t count, loff_t off)
 		return 0;
 	udev = to_usb_device (dev);
 
-	len = usb_string(udev, udev->descriptor.iSerialNumber, buf, PAGE_SIZE); 
+	len = usb_string(udev, udev->descriptor.iSerialNumber, buf, PAGE_SIZE);
+	if (len < 0)
+		return 0;
 	buf[len] = '\n';
-	buf[len+1] = 0x00;
+	buf[len+1] = 0;
 	return len+1;
 }
 static DEVICE_ATTR(serial,"serial",S_IRUGO,show_serial,NULL);
@@ -918,13 +924,13 @@ static void usb_find_drivers(struct usb_device *dev)
 	unsigned claimed = 0;
 
 	/* FIXME should get called for each new configuration not just the
-	 * first one for a device. switching configs (or altesettings) should
+	 * first one for a device. switching configs (or altsettings) should
 	 * undo driverfs and HCD state for the previous interfaces.
 	 */
 	for (ifnum = 0; ifnum < dev->actconfig->bNumInterfaces; ifnum++) {
 		struct usb_interface *interface = &dev->actconfig->interface[ifnum];
 		struct usb_interface_descriptor *desc = interface->altsetting;
-		
+
 		/* register this interface with driverfs */
 		interface->dev.parent = &dev->dev;
 		interface->dev.bus = &usb_bus_type;
@@ -1455,6 +1461,152 @@ int usb_new_device(struct usb_device *dev)
 }
 
 
+/**
+ * usb_buffer_alloc - allocate dma-consistent buffer for URB_NO_DMA_MAP
+ * @dev: device the buffer will be used with
+ * @size: requested buffer size
+ * @mem_flags: affect whether allocation may block
+ * @dma: used to return DMA address of buffer
+ *
+ * Return value is either null (indicating no buffer could be allocated), or
+ * the cpu-space pointer to a buffer that may be used to perform DMA to the
+ * specified device.  Such cpu-space buffers are returned along with the DMA
+ * address (through the pointer provided).
+ *
+ * These buffers are used with URB_NO_DMA_MAP set in urb->transfer_flags to
+ * avoid behaviors like using "DMA bounce buffers", or tying down I/O mapping
+ * hardware for long idle periods.  The implementation varies between
+ * platforms, depending on details of how DMA will work to this device.
+ *
+ * When the buffer is no longer used, free it with usb_buffer_free().
+ */
+void *usb_buffer_alloc (
+	struct usb_device *dev,
+	size_t size,
+	int mem_flags,
+	dma_addr_t *dma
+)
+{
+	if (!dev || !dev->bus || !dev->bus->op || !dev->bus->op->buffer_alloc)
+		return 0;
+	return dev->bus->op->buffer_alloc (dev->bus, size, mem_flags, dma);
+}
+
+/**
+ * usb_buffer_free - free memory allocated with usb_buffer_alloc()
+ * @dev: device the buffer was used with
+ * @size: requested buffer size
+ * @addr: CPU address of buffer
+ * @dma: DMA address of buffer
+ *
+ * This reclaims an I/O buffer, letting it be reused.  The memory must have
+ * been allocated using usb_buffer_alloc(), and the parameters must match
+ * those provided in that allocation request. 
+ */
+void usb_buffer_free (
+	struct usb_device *dev,
+	size_t size,
+	void *addr,
+	dma_addr_t dma
+)
+{
+	if (!dev || !dev->bus || !dev->bus->op || !dev->bus->op->buffer_free)
+	    	return;
+	dev->bus->op->buffer_free (dev->bus, size, addr, dma);
+}
+
+/**
+ * usb_buffer_map - create DMA mapping(s) for an urb
+ * @urb: urb whose transfer_buffer will be mapped
+ *
+ * Return value is either null (indicating no buffer could be mapped), or
+ * the parameter.  URB_NO_DMA_MAP is added to urb->transfer_flags if the
+ * operation succeeds.
+ *
+ * This call would normally be used for an urb which is reused, perhaps
+ * as the target of a large periodic transfer, with usb_buffer_dmasync()
+ * calls to synchronize memory and dma state.  It may not be used for
+ * control requests.
+ *
+ * Reverse the effect of this call with usb_buffer_unmap().
+ */
+struct urb *usb_buffer_map (struct urb *urb)
+{
+	struct usb_bus		*bus;
+	struct usb_operations	*op;
+
+	if (!urb
+			|| usb_pipecontrol (urb->pipe)
+			|| !urb->dev
+			|| !(bus = urb->dev->bus)
+			|| !(op = bus->op)
+			|| !op->buffer_map)
+		return 0;
+
+	if (op->buffer_map (bus,
+			urb->transfer_buffer,
+			&urb->transfer_dma,
+			urb->transfer_buffer_length,
+			usb_pipein (urb->pipe)
+				? USB_DIR_IN
+				: USB_DIR_OUT))
+		return 0;
+	urb->transfer_flags |= URB_NO_DMA_MAP;
+	return urb;
+}
+
+/**
+ * usb_buffer_dmasync - synchronize DMA and CPU view of buffer(s)
+ * @urb: urb whose transfer_buffer will be synchronized
+ */
+void usb_buffer_dmasync (struct urb *urb)
+{
+	struct usb_bus		*bus;
+	struct usb_operations	*op;
+
+	if (!urb
+			|| !(urb->transfer_flags & URB_NO_DMA_MAP)
+			|| !urb->dev
+			|| !(bus = urb->dev->bus)
+			|| !(op = bus->op)
+			|| !op->buffer_dmasync)
+		return;
+
+	op->buffer_dmasync (bus,
+			urb->transfer_dma,
+			urb->transfer_buffer_length,
+			usb_pipein (urb->pipe)
+				? USB_DIR_IN
+				: USB_DIR_OUT);
+}
+
+/**
+ * usb_buffer_unmap - free DMA mapping(s) for an urb
+ * @urb: urb whose transfer_buffer will be unmapped
+ *
+ * Reverses the effect of usb_buffer_map().
+ */
+void usb_buffer_unmap (struct urb *urb)
+{
+	struct usb_bus		*bus;
+	struct usb_operations	*op;
+
+	if (!urb
+			|| !(urb->transfer_flags & URB_NO_DMA_MAP)
+			|| !urb->dev
+			|| !(bus = urb->dev->bus)
+			|| !(op = bus->op)
+			|| !op->buffer_unmap)
+		return;
+
+	op->buffer_unmap (bus,
+			urb->transfer_dma,
+			urb->transfer_buffer_length,
+			usb_pipein (urb->pipe)
+				? USB_DIR_IN
+				: USB_DIR_OUT);
+}
+
 #ifdef CONFIG_PROC_FS
 struct list_head *usb_driver_get_list(void)
 {
@@ -1534,4 +1686,11 @@ EXPORT_SYMBOL(__usb_get_extra_descriptor);
 
 EXPORT_SYMBOL(usb_get_current_frame_number);
 
+EXPORT_SYMBOL (usb_buffer_alloc);
+EXPORT_SYMBOL (usb_buffer_free);
+
+EXPORT_SYMBOL (usb_buffer_map);
+EXPORT_SYMBOL (usb_buffer_dmasync);
+EXPORT_SYMBOL (usb_buffer_unmap);
+
 MODULE_LICENSE("GPL");

drivers/usb/host/ehci-dbg.c

 /*
- * Copyright (c) 2001 by David Brownell
+ * Copyright (c) 2001-2002 by David Brownell
  * 
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License as published by the
@@ -175,3 +175,215 @@ dbg_qh (char *label, struct ehci_hcd *ehci, struct ehci_qh *qh) {}
 		(status & PORT_CONNECT) ? " CONNECT" : "" \
 	    )
 
+#ifdef DEBUG
+
+#define speed_char(info1) ({ char tmp; \
+		switch (info1 & (3 << 12)) { \
+		case 0 << 12: tmp = 'f'; break; \
+		case 1 << 12: tmp = 'l'; break; \
+		case 2 << 12: tmp = 'h'; break; \
+		default: tmp = '?'; break; \
+		}; tmp; })
+
+static ssize_t
+show_async (struct device *dev, char *buf, size_t count, loff_t off)
+{
+	struct pci_dev		*pdev;
+	struct ehci_hcd		*ehci;
+	unsigned long		flags;
+	unsigned		temp, size;
+	char			*next;
+	struct ehci_qh		*qh;
+
+	if (off != 0)
+		return 0;
+
+	pdev = container_of (dev, struct pci_dev, dev);
+	ehci = container_of (pci_get_drvdata (pdev), struct ehci_hcd, hcd);
+	next = buf;
+	size = count;
+
+	/* dumps a snapshot of the async schedule.
+	 * usually empty except for long-term bulk reads, or head.
+	 * one QH per line, and TDs we know about
+	 */
+	spin_lock_irqsave (&ehci->lock, flags);
+	if (ehci->async) {
+		qh = ehci->async;
+		do {
+			u32			scratch;
+			struct list_head	*entry;
+			struct ehci_qtd		*td;
+
+			scratch = cpu_to_le32p (&qh->hw_info1);
+			temp = snprintf (next, size, "qh %p dev%d %cs ep%d",
+					qh, scratch & 0x007f,
+					speed_char (scratch),
+					(scratch >> 8) & 0x000f);
+			size -= temp;
+			next += temp;
+
+			list_for_each (entry, &qh->qtd_list) {
+				td = list_entry (entry, struct ehci_qtd,
+						qtd_list);
+				scratch = cpu_to_le32p (&td->hw_token);
+				temp = snprintf (next, size,
+						", td %p len=%d %s",
+						td, scratch >> 16,
+						({ char *tmp;
+						 switch ((scratch>>8)&0x03) {
+						 case 0: tmp = "out"; break;
+						 case 1: tmp = "in"; break;
+						 case 2: tmp = "setup"; break;
+						 default: tmp = "?"; break;
+						 } tmp;})
+						);
+				size -= temp;
+				next += temp;
+			}
+
+			temp = snprintf (next, size, "\n");
+			size -= temp;
+			next += temp;
+
+		} while ((qh = qh->qh_next.qh) != ehci->async);
+	}
+	spin_unlock_irqrestore (&ehci->lock, flags);
+
+	return count - size;
+}
+static DEVICE_ATTR (async, "sched-async", S_IRUSR, show_async, NULL);
+
+#define DBG_SCHED_LIMIT 64
+
+static ssize_t
+show_periodic (struct device *dev, char *buf, size_t count, loff_t off)
+{
+	struct pci_dev		*pdev;
+	struct ehci_hcd		*ehci;
+	unsigned long		flags;
+	union ehci_shadow	p, *seen;
+	unsigned		temp, size, seen_count;
+	char			*next;
+	unsigned		i, tag;
+
+	if (off != 0)
+		return 0;
+	if (!(seen = kmalloc (DBG_SCHED_LIMIT * sizeof *seen, SLAB_ATOMIC)))
+		return 0;
+	seen_count = 0;
+
+	pdev = container_of (dev, struct pci_dev, dev);
+	ehci = container_of (pci_get_drvdata (pdev), struct ehci_hcd, hcd);
+	next = buf;
+	size = count;
+
+	temp = snprintf (next, size, "size = %d\n", ehci->periodic_size);
+	size -= temp;
+	next += temp;
+
+	/* dump a snapshot of the periodic schedule.
+	 * iso changes, interrupt usually doesn't.
+	 */
+	spin_lock_irqsave (&ehci->lock, flags);
+	for (i = 0; i < ehci->periodic_size; i++) {
+		p = ehci->pshadow [i];
+		if (!p.ptr)
+			continue;
+		tag = Q_NEXT_TYPE (ehci->periodic [i]);
+
+		temp = snprintf (next, size, "%4d: ", i);
+		size -= temp;
+		next += temp;
+
+		do {
+			switch (tag) {
+			case Q_TYPE_QH:
+				temp = snprintf (next, size, " intr-%d %p",
+						p.qh->period, p.qh);
+				size -= temp;
+				next += temp;
+				for (temp = 0; temp < seen_count; temp++) {
+					if (seen [temp].ptr == p.ptr)
+						break;
+				}
+				/* show more info the first time around */
+				if (temp == seen_count) {
+					u32	scratch = cpu_to_le32p (
+							&p.qh->hw_info1);
+
+					temp = snprintf (next, size,
+						" (%cs dev%d ep%d)",
+						speed_char (scratch),
+						scratch & 0x007f,
+						(scratch >> 8) & 0x000f);
+
+					/* FIXME TDs too */
+
+					if (seen_count < DBG_SCHED_LIMIT)
+						seen [seen_count++].qh = p.qh;
+				} else
+					temp = 0;
+				tag = Q_NEXT_TYPE (p.qh->hw_next);
+				p = p.qh->qh_next;
+				break;
+			case Q_TYPE_FSTN:
+				temp = snprintf (next, size,
+					" fstn-%8x/%p", p.fstn->hw_prev,
+					p.fstn);
+				tag = Q_NEXT_TYPE (p.fstn->hw_next);
+				p = p.fstn->fstn_next;
+				break;
+			case Q_TYPE_ITD:
+				temp = snprintf (next, size,
+					" itd/%p", p.itd);
+				tag = Q_NEXT_TYPE (p.itd->hw_next);
+				p = p.itd->itd_next;
+				break;
+			case Q_TYPE_SITD:
+				temp = snprintf (next, size,
+					" sitd/%p", p.sitd);
+				tag = Q_NEXT_TYPE (p.sitd->hw_next);
+				p = p.sitd->sitd_next;
+				break;
+			}
+			size -= temp;
+			next += temp;
+		} while (p.ptr);
+
+		temp = snprintf (next, size, "\n");
+		size -= temp;
+		next += temp;
+	}
+	spin_unlock_irqrestore (&ehci->lock, flags);
+	kfree (seen);
+
+	return count - size;
+}
+static DEVICE_ATTR (periodic, "sched-periodic", S_IRUSR, show_periodic, NULL);
+
+#undef DBG_SCHED_LIMIT
+
+static inline void create_debug_files (struct ehci_hcd *bus)
+{
+	device_create_file (&bus->hcd.pdev->dev, &dev_attr_async);
+	device_create_file (&bus->hcd.pdev->dev, &dev_attr_periodic);
+}
+
+static inline void remove_debug_files (struct ehci_hcd *bus)
+{
+	device_remove_file (&bus->hcd.pdev->dev, &dev_attr_async);
+	device_remove_file (&bus->hcd.pdev->dev, &dev_attr_periodic);
+}
+
+#else /* DEBUG */
+
+static inline void create_debug_files (struct ehci_hcd *bus)
+{
+}
+
+static inline void remove_debug_files (struct ehci_hcd *bus)
+{
+}
+
+#endif /* DEBUG */

drivers/usb/host/ehci-hcd.c

@@ -65,6 +65,8 @@
  *
  * HISTORY:
  *
+ * 2002-08-06	Handling for bulk and interrupt transfers is mostly shared;
+ *	only scheduling is different, no arbitrary limitations.
  * 2002-07-25	Sanity check PCI reads, mostly for better cardbus support,
  * 	clean up HC run state handshaking.
  * 2002-05-24	Preliminary FS/LS interrupts, using scheduling shortcuts
@@ -85,7 +87,7 @@
  * 2001-June	Works with usb-storage and NEC EHCI on 2.4
  */
 
-#define DRIVER_VERSION "2002-Jul-25"
+#define DRIVER_VERSION "2002-Aug-06"
 #define DRIVER_AUTHOR "David Brownell"
 #define DRIVER_DESC "USB 2.0 'Enhanced' Host Controller (EHCI) Driver"
 
@@ -93,6 +95,8 @@
 // #define EHCI_VERBOSE_DEBUG
 // #define have_split_iso
 
+#define INTR_AUTOMAGIC		/* to be removed later in 2.5 */
+
 /* magic numbers that can affect system performance */
 #define	EHCI_TUNE_CERR		3	/* 0-3 qtd retries; 0 == don't stop */
 #define	EHCI_TUNE_RL_HS		0	/* nak throttle; see 4.9 */
@@ -376,6 +380,8 @@ static int ehci_start (struct usb_hcd *hcd)
 		return -ENOMEM;
 	}
 
+	create_debug_files (ehci);
+
 	/*
 	 * Start, enabling full USB 2.0 functionality ... usb 1.1 devices
 	 * are explicitly handed to companion controller(s), so no TT is
@@ -429,6 +435,8 @@ static void ehci_stop (struct usb_hcd *hcd)
 		ehci_ready (ehci);
 	ehci_reset (ehci);
 
+	remove_debug_files (ehci);
+
 	/* root hub is shut down separately (first, when possible) */
 	tasklet_disable (&ehci->tasklet);
 	ehci_tasklet ((unsigned long) ehci);
@@ -614,7 +622,8 @@ static void ehci_irq (struct usb_hcd *hcd)
  *
  * hcd-specific init for hcpriv hasn't been done yet
  *
- * NOTE:  EHCI queues control and bulk requests transparently, like OHCI.
+ * NOTE:  control, bulk, and interrupt share the same code to append TDs
+ * to a (possibly active) QH, and the same QH scanning code.
  */
 static int ehci_urb_enqueue (
 	struct usb_hcd	*hcd,
@@ -626,10 +635,11 @@ static int ehci_urb_enqueue (
 
 	urb->transfer_flags &= ~EHCI_STATE_UNLINK;
 	INIT_LIST_HEAD (&qtd_list);
-	switch (usb_pipetype (urb->pipe)) {
 
-	case PIPE_CONTROL:
-	case PIPE_BULK:
+	switch (usb_pipetype (urb->pipe)) {
+	// case PIPE_CONTROL:
+	// case PIPE_BULK:
+	default:
 		if (!qh_urb_transaction (ehci, urb, &qtd_list, mem_flags))
 			return -ENOMEM;
 		return submit_async (ehci, urb, &qtd_list, mem_flags);
@@ -649,9 +659,6 @@ static int ehci_urb_enqueue (
 		dbg ("no split iso support yet");
 		return -ENOSYS;
 #endif /* have_split_iso */
-
-	default:	/* can't happen */
-		return -ENOSYS;
 	}
 }
 
@@ -665,15 +672,16 @@ static int ehci_urb_dequeue (struct usb_hcd *hcd, struct urb *urb)
 	struct ehci_qh		*qh = (struct ehci_qh *) urb->hcpriv;
 	unsigned long		flags;
 
-	dbg ("%s urb_dequeue %p qh state %d",
-		hcd->self.bus_name, urb, qh->qh_state);
+	dbg ("%s urb_dequeue %p qh %p state %d",
+		hcd->self.bus_name, urb, qh, qh->qh_state);
 
 	switch (usb_pipetype (urb->pipe)) {
-	case PIPE_CONTROL:
-	case PIPE_BULK:
+	// case PIPE_CONTROL:
+	// case PIPE_BULK:
+	default:
 		spin_lock_irqsave (&ehci->lock, flags);
 		if (ehci->reclaim) {
-dbg ("dq: reclaim busy, %s", RUN_CONTEXT);
+			dbg ("dq: reclaim busy, %s", RUN_CONTEXT);
 			if (in_interrupt ()) {
 				spin_unlock_irqrestore (&ehci->lock, flags);
 				return -EAGAIN;
@@ -683,28 +691,43 @@ dbg ("dq: reclaim busy, %s", RUN_CONTEXT);
 					&& ehci->hcd.state != USB_STATE_HALT
 					) {
 				spin_unlock_irqrestore (&ehci->lock, flags);
-// yeech ... this could spin for up to two frames!
-dbg ("wait for dequeue: state %d, reclaim %p, hcd state %d",
-    qh->qh_state, ehci->reclaim, ehci->hcd.state
-);
-				udelay (100);
+				/* let pending unlinks complete */
+				wait_ms (1);
 				spin_lock_irqsave (&ehci->lock, flags);
 			}
 		}
 		if (qh->qh_state == QH_STATE_LINKED)
 			start_unlink_async (ehci, qh);
 		spin_unlock_irqrestore (&ehci->lock, flags);
-		return 0;
+		break;
 
 	case PIPE_INTERRUPT:
-		intr_deschedule (ehci, urb->start_frame, qh,
-			(urb->dev->speed == USB_SPEED_HIGH)
-			    ? urb->interval
-			    : (urb->interval << 3));
-		if (ehci->hcd.state == USB_STATE_HALT)
-			urb->status = -ESHUTDOWN;
-		qh_completions (ehci, qh, 1);
-		return 0;
+		if (qh->qh_state == QH_STATE_LINKED) {
+			/* messy, can spin or block a microframe ... */
+			intr_deschedule (ehci, qh, 1);
+			/* qh_state == IDLE */
+		}
+		qh_completions (ehci, qh);
+
+		/* reschedule QH iff another request is queued */
+		if (!list_empty (&qh->qtd_list)
+				&& HCD_IS_RUNNING (ehci->hcd.state)) {
+			int status;
+
+			spin_lock_irqsave (&ehci->lock, flags);
+			status = qh_schedule (ehci, qh);
+			spin_unlock_irqrestore (&ehci->lock, flags);
+
+			if (status != 0) {
+				// shouldn't happen often, but ...
+				// FIXME kill those tds' urbs
+				err ("can't reschedule qh %p, err %d",
+					qh, status);
+			}
+			return status;
+		}
+
+		break;
 
 	case PIPE_ISOCHRONOUS:
 		// itd or sitd ...
@@ -712,9 +735,9 @@ dbg ("wait for dequeue: state %d, reclaim %p, hcd state %d",
 		// wait till next completion, do it then.
 		// completion irqs can wait up to 1024 msec,
 		urb->transfer_flags |= EHCI_STATE_UNLINK;
-		return 0;
+		break;
 	}
-	return -EINVAL;
+	return 0;
 }
 
 /*-------------------------------------------------------------------------*/
@@ -728,6 +751,7 @@ static void ehci_free_config (struct usb_hcd *hcd, struct usb_device *udev)
 	int			i;
 	unsigned long		flags;
 
+	/* ASSERT:  no requests/urbs are still linked (so no TDs) */
 	/* ASSERT:  nobody can be submitting urbs for this any more */
 
 	dbg ("%s: free_config devnum %d", hcd->self.bus_name, udev->devnum);
@@ -736,34 +760,57 @@ static void ehci_free_config (struct usb_hcd *hcd, struct usb_device *udev)
 	for (i = 0; i < 32; i++) {
 		if (dev->ep [i]) {
 			struct ehci_qh		*qh;
+			char			*why;
 
 			/* dev->ep never has ITDs or SITDs */
 			qh = (struct ehci_qh *) dev->ep [i];
-			vdbg ("free_config, ep 0x%02x qh %p", i, qh);
-			if (!list_empty (&qh->qtd_list)) {
-				dbg ("ep 0x%02x qh %p not empty!", i, qh);
+
+			/* detect/report non-recoverable errors */
+			if (in_interrupt ()) 
+				why = "disconnect() didn't";
+			else if ((qh->hw_info2 & cpu_to_le32 (0xffff)) != 0
+					&& qh->qh_state != QH_STATE_IDLE)
+				why = "(active periodic)";
+			else
+				why = 0;
+			if (why) {
+				err ("dev %s-%s ep %d-%s error: %s",
+					hcd->self.bus_name, udev->devpath,
+					i & 0xf, (i & 0x10) ? "IN" : "OUT",
+					why);
 				BUG ();
 			}
-			dev->ep [i] = 0;
 
-			/* wait_ms() won't spin here -- we're a thread */
+			dev->ep [i] = 0;
+			if (qh->qh_state == QH_STATE_IDLE)
+				goto idle;
+			dbg ("free_config, async ep 0x%02x qh %p", i, qh);
+
+			/* scan_async() empties the ring as it does its work,
+			 * using IAA, but doesn't (yet?) turn it off.  if it
+			 * doesn't empty this qh, likely it's the last entry.
+			 */
 			while (qh->qh_state == QH_STATE_LINKED
 					&& ehci->reclaim
 					&& ehci->hcd.state != USB_STATE_HALT
 					) {
 				spin_unlock_irqrestore (&ehci->lock, flags);
+				/* wait_ms() won't spin, we're a thread;
+				 * and we know IRQ+tasklet can progress
+				 */
 				wait_ms (1);
 				spin_lock_irqsave (&ehci->lock, flags);
 			}
-			if (qh->qh_state == QH_STATE_LINKED) {
+			if (qh->qh_state == QH_STATE_LINKED)
 				start_unlink_async (ehci, qh);
-				while (qh->qh_state != QH_STATE_IDLE) {
-					spin_unlock_irqrestore (&ehci->lock,
-						flags);
-					wait_ms (1);
-					spin_lock_irqsave (&ehci->lock, flags);
-				}
+			while (qh->qh_state != QH_STATE_IDLE
+					&& ehci->hcd.state != USB_STATE_HALT) {
+				spin_unlock_irqrestore (&ehci->lock,
+					flags);
+				wait_ms (1);
+				spin_lock_irqsave (&ehci->lock, flags);
 			}
+idle:
 			qh_put (ehci, qh);
 		}
 	}

drivers/usb/host/ehci-q.c

@@ -47,9 +47,11 @@ static int
 qtd_fill (struct ehci_qtd *qtd, dma_addr_t buf, size_t len, int token)
 {
 	int	i, count;
+	u64	addr = buf;
 
 	/* one buffer entry per 4K ... first might be short or unaligned */
-	qtd->hw_buf [0] = cpu_to_le32 (buf);
+	qtd->hw_buf [0] = cpu_to_le32 ((u32)addr);
+	qtd->hw_buf_hi [0] = cpu_to_le32 ((u32)(addr >> 32));
 	count = 0x1000 - (buf & 0x0fff);	/* rest of that page */
 	if (likely (len < count))		/* ... iff needed */
 		count = len;
@@ -59,7 +61,7 @@ qtd_fill (struct ehci_qtd *qtd, dma_addr_t buf, size_t len, int token)
 
 		/* per-qtd limit: from 16K to 20K (best alignment) */
 		for (i = 1; count < len && i < 5; i++) {
-			u64	addr = buf;
+			addr = buf;
 			qtd->hw_buf [i] = cpu_to_le32 ((u32)addr);
 			qtd->hw_buf_hi [i] = cpu_to_le32 ((u32)(addr >> 32));
 			buf += 0x1000;
@@ -157,30 +159,6 @@ static inline void qtd_copy_status (struct urb *urb, size_t length, u32 token)
 	}
 }
 
-static void ehci_urb_complete (
-	struct ehci_hcd		*ehci,
-	dma_addr_t		addr,
-	struct urb		*urb
-) {
-	if (urb->transfer_buffer_length && usb_pipein (urb->pipe))
-		pci_dma_sync_single (ehci->hcd.pdev, addr,
-			urb->transfer_buffer_length,
-			PCI_DMA_FROMDEVICE);
-
-	/* cleanse status if we saw no error */
-	if (likely (urb->status == -EINPROGRESS)) {
-		if (urb->actual_length != urb->transfer_buffer_length
-				&& (urb->transfer_flags & URB_SHORT_NOT_OK))
-			urb->status = -EREMOTEIO;
-		else
-			urb->status = 0;
-	}
-
-	/* only report unlinks once */
-	if (likely (urb->status != -ENOENT && urb->status != -ENOTCONN))
-		urb->complete (urb);
-}
-
 /* urb->lock ignored from here on (hcd is done with urb) */
 
 static void ehci_urb_done (
@@ -188,6 +166,11 @@ static void ehci_urb_done (
 	dma_addr_t		addr,
 	struct urb		*urb
 ) {
+#ifdef	INTR_AUTOMAGIC
+	struct urb		*resubmit = 0;
+	struct usb_device	*dev = 0;
+#endif
+
 	if (urb->transfer_buffer_length)
 		pci_unmap_single (ehci->hcd.pdev,
 			addr,
@@ -196,7 +179,23 @@ static void ehci_urb_done (
 			    ? PCI_DMA_FROMDEVICE
 			    : PCI_DMA_TODEVICE);
 	if (likely (urb->hcpriv != 0)) {
-		qh_put (ehci, (struct ehci_qh *) urb->hcpriv);
+		struct ehci_qh	*qh = (struct ehci_qh *) urb->hcpriv;
+
+		/* S-mask in a QH means it's an interrupt urb */
+		if ((qh->hw_info2 & cpu_to_le32 (0x00ff)) != 0) {
+
+			/* ... update hc-wide periodic stats (for usbfs) */
+			ehci->hcd.self.bandwidth_int_reqs--;
+
+#ifdef	INTR_AUTOMAGIC
+			if (!((urb->status == -ENOENT)
+					|| (urb->status == -ECONNRESET))) {
+				resubmit = usb_get_urb (urb);
+				dev = urb->dev;
+			}
+#endif
+		}
+		qh_put (ehci, qh);
 		urb->hcpriv = 0;
 	}
 
@@ -208,33 +207,46 @@ static void ehci_urb_done (
 			urb->status = 0;
 	}
 
-	/* hand off urb ownership */
 	usb_hcd_giveback_urb (&ehci->hcd, urb);
+
+#ifdef	INTR_AUTOMAGIC
+	if (resubmit && ((urb->status == -ENOENT)
+				|| (urb->status == -ECONNRESET))) {
+		usb_put_urb (resubmit);
+		resubmit = 0;
+	}
+	// device drivers will soon be doing something like this
+	if (resubmit) {
+		int	status;
+
+		resubmit->dev = dev;
+		status = usb_submit_urb (resubmit, SLAB_KERNEL);
+		if (status != 0)
+			err ("can't resubmit interrupt urb %p: status %d",
+					resubmit, status);
+		usb_put_urb (resubmit);
+	}
+#endif
 }
 
 
 /*
  * Process completed qtds for a qh, issuing completions if needed.
- * When freeing:  frees qtds, unmaps buf, returns URB to driver.
- * When not freeing (queued periodic qh):  retain qtds, mapping, and urb.
+ * Frees qtds, unmaps buf, returns URB to driver.
  * Races up to qh->hw_current; returns number of urb completions.
  */
-static int
-qh_completions (
-	struct ehci_hcd		*ehci,
-	struct ehci_qh		*qh,
-	int			freeing
-) {
+static void
+qh_completions (struct ehci_hcd *ehci, struct ehci_qh *qh)
+{
 	struct ehci_qtd		*qtd, *last;
 	struct list_head	*next, *qtd_list = &qh->qtd_list;
 	int			unlink = 0, halted = 0;
 	unsigned long		flags;
-	int			retval = 0;
 
 	spin_lock_irqsave (&ehci->lock, flags);
 	if (unlikely (list_empty (qtd_list))) {
 		spin_unlock_irqrestore (&ehci->lock, flags);
-		return retval;
+		return;
 	}
 
 	/* scan QTDs till end of list, or we reach an active one */
@@ -251,14 +263,8 @@ qh_completions (
 			if (likely (last->urb != urb)) {
 				/* complete() can reenter this HCD */
 				spin_unlock_irqrestore (&ehci->lock, flags);
-				if (likely (freeing != 0))
-					ehci_urb_done (ehci, last->buf_dma,
-						last->urb);
-				else
-					ehci_urb_complete (ehci, last->buf_dma,
-						last->urb);
+				ehci_urb_done (ehci, last->buf_dma, last->urb);
 				spin_lock_irqsave (&ehci->lock, flags);
-				retval++;
 			}
 
 			/* qh overlays can have HC's old cached copies of
@@ -270,8 +276,7 @@ qh_completions (
 				qh->hw_qtd_next = last->hw_next;
 			}
 
-			if (likely (freeing != 0))
-				ehci_qtd_free (ehci, last);
+			ehci_qtd_free (ehci, last);
 			last = 0;
 		}
 		next = qtd->qtd_list.next;
@@ -288,7 +293,7 @@ qh_completions (
 
 		/* fault: unlink the rest, since this qtd saw an error? */
 		if (unlikely ((token & QTD_STS_HALT) != 0)) {
-			freeing = unlink = 1;
+			unlink = 1;
 			/* status copied below */
 
 		/* QH halts only because of fault (above) or unlink (here). */
@@ -296,13 +301,14 @@ qh_completions (
 
 			/* unlinking everything because of HC shutdown? */
 			if (ehci->hcd.state == USB_STATE_HALT) {
-				freeing = unlink = 1;
+				unlink = 1;
 
 			/* explicit unlink, maybe starting here? */
 			} else if (qh->qh_state == QH_STATE_IDLE
 					&& (urb->status == -ECONNRESET
+						|| urb->status == -ESHUTDOWN
 						|| urb->status == -ENOENT)) {
-				freeing = unlink = 1;
+				unlink = 1;
 
 			/* QH halted to unlink urbs _after_ this?  */
 			} else if (!unlink && (token & QTD_STS_ACTIVE) != 0) {
@@ -312,7 +318,7 @@ qh_completions (
 
 			/* unlink the rest?  once we start unlinking, after
 			 * a fault or explicit unlink, we unlink all later
-			 * urbs.  usb spec requires that.
+			 * urbs.  usb spec requires that for faults...
 			 */
 			if (unlink && urb->status == -EINPROGRESS)
 				urb->status = -ECONNRESET;
@@ -330,31 +336,7 @@ qh_completions (
 		qtd_copy_status (urb, qtd->length, token);
 		spin_unlock (&urb->lock);
 
-		/*
-		 * NOTE:  this won't work right with interrupt urbs that
-		 * need multiple qtds ... only the first scan of qh->qtd_list
-		 * starts at the right qtd, yet multiple scans could happen
-		 * for transfers that are scheduled across multiple uframes. 
-		 * (Such schedules are not currently allowed!)
-		 */
-		if (likely (freeing != 0))
-			list_del (&qtd->qtd_list);
-		else {
-			/* restore everything the HC could change
-			 * from an interrupt QTD
-			 */
-			qtd->hw_token = (qtd->hw_token
-					& __constant_cpu_to_le32 (0x8300))
-				| cpu_to_le32 (qtd->length << 16)
-				| __constant_cpu_to_le32 (QTD_STS_ACTIVE
-					| (EHCI_TUNE_CERR << 10));
-			qtd->hw_buf [0] &= ~__constant_cpu_to_le32 (0x0fff);
-
-			/* this offset, and the length above,
-			 * are likely wrong on QTDs #2..N
-			 */
-			qtd->hw_buf [0] |= cpu_to_le32 (0x0fff & qtd->buf_dma);
-		}
+		list_del (&qtd->qtd_list);
 
 #if 0
 		if (urb->status == -EINPROGRESS)
@@ -382,14 +364,9 @@ qh_completions (
 
 	/* last urb's completion might still need calling */
 	if (likely (last != 0)) {
-		if (likely (freeing != 0)) {
-			ehci_urb_done (ehci, last->buf_dma, last->urb);
-			ehci_qtd_free (ehci, last);
-		} else
-			ehci_urb_complete (ehci, last->buf_dma, last->urb);
-		retval++;
+		ehci_urb_done (ehci, last->buf_dma, last->urb);
+		ehci_qtd_free (ehci, last);
 	}
-	return retval;
 }
 
 /*-------------------------------------------------------------------------*/
@@ -450,6 +427,7 @@ qh_urb_transaction (
 	struct ehci_qtd		*qtd, *qtd_prev;
 	dma_addr_t		buf, map_buf;
 	int			len, maxpacket;
+	int			is_input;
 	u32			token;
 
 	/*
@@ -495,10 +473,11 @@ qh_urb_transaction (
 	 * data transfer stage:  buffer setup
 	 */
 	len = urb->transfer_buffer_length;
+	is_input = usb_pipein (urb->pipe);
 	if (likely (len > 0)) {
 		buf = map_buf = pci_map_single (ehci->hcd.pdev,
 			urb->transfer_buffer, len,
-			usb_pipein (urb->pipe)
+			is_input
 			    ? PCI_DMA_FROMDEVICE
 			    : PCI_DMA_TODEVICE);
 		if (unlikely (!buf))
@@ -506,12 +485,11 @@ qh_urb_transaction (
 	} else
 		buf = map_buf = 0;
 
-	if (!buf || usb_pipein (urb->pipe))
+	if (!buf || is_input)
 		token |= (1 /* "in" */ << 8);
 	/* else it's already initted to "out" pid (0 << 8) */
 
-	maxpacket = usb_maxpacket (urb->dev, urb->pipe,
-			usb_pipeout (urb->pipe));
+	maxpacket = usb_maxpacket (urb->dev, urb->pipe, !is_input) & 0x03ff;
 
 	/*
 	 * buffer gets wrapped in one or more qtds;
@@ -607,6 +585,11 @@ clear_toggle (struct usb_device *udev, int ep, int is_out, struct ehci_qh *qh)
 // That'd mean updating how usbcore talks to HCDs. (2.5?)
 
 
+// high bandwidth multiplier, as encoded in highspeed endpoint descriptors
+#define hb_mult(wMaxPacketSize) (1 + (((wMaxPacketSize) >> 11) & 0x03))
+// ... and packet size, for any kind of endpoint descriptor
+#define max_packet(wMaxPacketSize) ((wMaxPacketSize) & 0x03ff)
+
 /*
  * Each QH holds a qtd list; a QH is used for everything except iso.
  *
@@ -624,6 +607,8 @@ ehci_qh_make (
 ) {
 	struct ehci_qh		*qh = ehci_qh_alloc (ehci, flags);
 	u32			info1 = 0, info2 = 0;
+	int			is_input, type;
+	int			maxp = 0;
 
 	if (!qh)
 		return qh;
@@ -634,6 +619,53 @@ ehci_qh_make (
 	info1 |= usb_pipeendpoint (urb->pipe) << 8;
 	info1 |= usb_pipedevice (urb->pipe) << 0;
 
+	is_input = usb_pipein (urb->pipe);
+	type = usb_pipetype (urb->pipe);
+	maxp = usb_maxpacket (urb->dev, urb->pipe, !is_input);
+
+	/* Compute interrupt scheduling parameters just once, and save.
+	 * - allowing for high bandwidth, how many nsec/uframe are used?
+	 * - split transactions need a second CSPLIT uframe; same question
+	 * - splits also need a schedule gap (for full/low speed I/O)
+	 * - qh has a polling interval
+	 *
+	 * For control/bulk requests, the HC or TT handles these.
+	 */
+	if (type == PIPE_INTERRUPT) {
+		qh->usecs = usb_calc_bus_time (USB_SPEED_HIGH, is_input, 0,
+				hb_mult (maxp) * max_packet (maxp));
+		qh->start = NO_FRAME;
+
+		if (urb->dev->speed == USB_SPEED_HIGH) {
+			qh->c_usecs = 0;
+			qh->gap_uf = 0;
+
+			/* FIXME handle HS periods of less than 1 frame. */
+			qh->period = urb->interval >> 3;
+			if (qh->period < 1) {
+				dbg ("intr period %d uframes, NYET!",
+						urb->interval);
+				qh = 0;
+				goto done;
+			}
+		} else {
+			/* gap is f(FS/LS transfer times) */
+			qh->gap_uf = 1 + usb_calc_bus_time (urb->dev->speed,
+					is_input, 0, maxp) / (125 * 1000);
+
+			/* FIXME this just approximates SPLIT/CSPLIT times */
+			if (is_input) {		// SPLIT, gap, CSPLIT+DATA
+				qh->c_usecs = qh->usecs + HS_USECS (0);
+				qh->usecs = HS_USECS (1);
+			} else {		// SPLIT+DATA, gap, CSPLIT
+				qh->usecs += HS_USECS (1);
+				qh->c_usecs = HS_USECS (0);
+			}
+
+			qh->period = urb->interval;
+		}
+	}
+
 	/* using TT? */
 	switch (urb->dev->speed) {
 	case USB_SPEED_LOW:
@@ -643,67 +675,63 @@ ehci_qh_make (
 	case USB_SPEED_FULL:
 		/* EPS 0 means "full" */
 		info1 |= (EHCI_TUNE_RL_TT << 28);
-		if (usb_pipecontrol (urb->pipe)) {
+		if (type == PIPE_CONTROL) {
 			info1 |= (1 << 27);	/* for TT */
 			info1 |= 1 << 14;	/* toggle from qtd */
 		}
-		info1 |= usb_maxpacket (urb->dev, urb->pipe,
-					usb_pipeout (urb->pipe)) << 16;
+		info1 |= maxp << 16;
 
 		info2 |= (EHCI_TUNE_MULT_TT << 30);
 		info2 |= urb->dev->ttport << 23;
 		info2 |= urb->dev->tt->hub->devnum << 16;
 
-		/* NOTE:  if (usb_pipeint (urb->pipe)) { scheduler sets c-mask }
-		 * ... and a 0.96 scheduler might use FSTN nodes too
-		 */
+		/* NOTE:  if (PIPE_INTERRUPT) { scheduler sets c-mask } */
+
 		break;
 
 	case USB_SPEED_HIGH:		/* no TT involved */
 		info1 |= (2 << 12);	/* EPS "high" */
 		info1 |= (EHCI_TUNE_RL_HS << 28);
-		if (usb_pipecontrol (urb->pipe)) {
+		if (type == PIPE_CONTROL) {
 			info1 |= 64 << 16;	/* usb2 fixed maxpacket */
 			info1 |= 1 << 14;	/* toggle from qtd */
 			info2 |= (EHCI_TUNE_MULT_HS << 30);
-		} else if (usb_pipebulk (urb->pipe)) {
+		} else if (type == PIPE_BULK) {
 			info1 |= 512 << 16;	/* usb2 fixed maxpacket */
 			info2 |= (EHCI_TUNE_MULT_HS << 30);
-		} else {
-			u32	temp;
-			temp = usb_maxpacket (urb->dev, urb->pipe,
-						usb_pipeout (urb->pipe));
-			info1 |= (temp & 0x3ff) << 16;	/* maxpacket */
-			/* HS intr can be "high bandwidth" */
-			temp = 1 + ((temp >> 11) & 0x03);
-			info2 |= temp << 30;		/* mult */
+		} else {		/* PIPE_INTERRUPT */
+			info1 |= max_packet (maxp) << 16;
+			info2 |= hb_mult (maxp) << 30;
 		}
 		break;
-	default:
 #ifdef DEBUG
+	default:
 		BUG ();
 #endif
 	}
 
-	/* NOTE:  if (usb_pipeint (urb->pipe)) { scheduler sets s-mask } */
+	/* NOTE:  if (PIPE_INTERRUPT) { scheduler sets s-mask } */
 
 	qh->qh_state = QH_STATE_IDLE;
 	qh->hw_info1 = cpu_to_le32 (info1);
 	qh->hw_info2 = cpu_to_le32 (info2);
 
 	/* initialize sw and hw queues with these qtds */
-	list_splice (qtd_list, &qh->qtd_list);
-	qh_update (qh, list_entry (qtd_list->next, struct ehci_qtd, qtd_list));
+	if (!list_empty (qtd_list)) {
+		list_splice (qtd_list, &qh->qtd_list);
+		qh_update (qh, list_entry (qtd_list->next, struct ehci_qtd, qtd_list));
+	} else {
+		qh->hw_qtd_next = qh->hw_alt_next = EHCI_LIST_END;
+	}
 
 	/* initialize data toggle state */
-	if (!usb_pipecontrol (urb->pipe))
-		clear_toggle (urb->dev,
-			usb_pipeendpoint (urb->pipe),
-			usb_pipeout (urb->pipe),
-			qh);
+	clear_toggle (urb->dev, usb_pipeendpoint (urb->pipe), !is_input, qh);
 
+done:
 	return qh;
 }
+#undef hb_mult
+#undef hb_packet
 
 /*-------------------------------------------------------------------------*/
 
@@ -745,50 +773,48 @@ static void qh_link_async (struct ehci_hcd *ehci, struct ehci_qh *qh)
 
 /*-------------------------------------------------------------------------*/
 
-static int
-submit_async (
+/*
+ * For control/bulk/interrupt, return QH with these TDs appended.
+ * Allocates and initializes the QH if necessary.
+ * Returns null if it can't allocate a QH it needs to.
+ * If the QH has TDs (urbs) already, that's great.
+ */
+static struct ehci_qh *qh_append_tds (
 	struct ehci_hcd		*ehci,
 	struct urb		*urb,
 	struct list_head	*qtd_list,
-	int			mem_flags
-) {
-	struct ehci_qtd		*qtd;
-	struct hcd_dev		*dev;
-	int			epnum;
-	unsigned long		flags;
+	int			epnum,
+	void			**ptr
+)
+{
 	struct ehci_qh		*qh = 0;
 
-	qtd = list_entry (qtd_list->next, struct ehci_qtd, qtd_list);
-	dev = (struct hcd_dev *)urb->dev->hcpriv;
-	epnum = usb_pipeendpoint (urb->pipe);
-	if (usb_pipein (urb->pipe))
-		epnum |= 0x10;
-
-	vdbg ("%s: submit_async urb %p len %d ep %d-%s qtd %p [qh %p]",
-		ehci->hcd.self.bus_name, urb, urb->transfer_buffer_length,
-		epnum & 0x0f, (epnum & 0x10) ? "in" : "out",
-		qtd, dev ? dev->ep [epnum] : (void *)~0);
-
-	spin_lock_irqsave (&ehci->lock, flags);
-
-	qh = (struct ehci_qh *) dev->ep [epnum];
+	qh = (struct ehci_qh *) *ptr;
 	if (likely (qh != 0)) {
-		u32	hw_next = QTD_NEXT (qtd->qtd_dma);
+		struct ehci_qtd	*qtd;
+
+		if (unlikely (list_empty (qtd_list)))
+			qtd = 0;
+		else
+			qtd = list_entry (qtd_list->next, struct ehci_qtd,
+					qtd_list);
 
 		/* maybe patch the qh used for set_address */
 		if (unlikely (epnum == 0
 				&& le32_to_cpu (qh->hw_info1 & 0x7f) == 0))
 			qh->hw_info1 |= cpu_to_le32 (usb_pipedevice(urb->pipe));
 
-		/* is an URB is queued to this qh already? */
-		if (unlikely (!list_empty (&qh->qtd_list))) {
+		/* append to tds already queued to this qh? */
+		if (unlikely (!list_empty (&qh->qtd_list) && qtd)) {
 			struct ehci_qtd		*last_qtd;
 			int			short_rx = 0;
+			u32			hw_next;
 
 			/* update the last qtd's "next" pointer */
 			// dbg_qh ("non-empty qh", ehci, qh);
 			last_qtd = list_entry (qh->qtd_list.prev,
 					struct ehci_qtd, qtd_list);
+			hw_next = QTD_NEXT (qtd->qtd_dma);
 			last_qtd->hw_next = hw_next;
 
 			/* previous urb allows short rx? maybe optimize. */
@@ -803,6 +829,7 @@ submit_async (
 			 * Interrupt code must cope with case of HC having it
 			 * cached, and clobbering these updates.
 			 * ... complicates getting rid of extra interrupts!
+			 * (Or:  use dummy td, so cache always stays valid.)
 			 */
 			if (qh->hw_current == cpu_to_le32 (last_qtd->qtd_dma)) {
 				wmb ();
@@ -822,31 +849,61 @@ submit_async (
 			 */
 
 			/* usb_clear_halt() means qh data toggle gets reset */
-			if (usb_pipebulk (urb->pipe)
-					&& unlikely (!usb_gettoggle (urb->dev,
+			if (unlikely (!usb_gettoggle (urb->dev,
 						(epnum & 0x0f),
 						!(epnum & 0x10)))) {
 				clear_toggle (urb->dev,
 					epnum & 0x0f, !(epnum & 0x10), qh);
 			}
-			qh_update (qh, qtd);
+			if (qtd)
+				qh_update (qh, qtd);
 		}
 		list_splice (qtd_list, qh->qtd_list.prev);
 
 	} else {
 		/* can't sleep here, we have ehci->lock... */
 		qh = ehci_qh_make (ehci, urb, qtd_list, SLAB_ATOMIC);
-		if (likely (qh != 0)) {
-			// dbg_qh ("new qh", ehci, qh);
-			dev->ep [epnum] = qh;
-		}
+		// if (qh) dbg_qh ("new qh", ehci, qh);
+		*ptr = qh;
 	}
+	if (qh)
+		urb->hcpriv = qh_get (qh);
+	return qh;
+}
+
+/*-------------------------------------------------------------------------*/
+
+static int
+submit_async (
+	struct ehci_hcd		*ehci,
+	struct urb		*urb,
+	struct list_head	*qtd_list,
+	int			mem_flags
+) {
+	struct ehci_qtd		*qtd;
+	struct hcd_dev		*dev;
+	int			epnum;
+	unsigned long		flags;
+	struct ehci_qh		*qh = 0;
+
+	qtd = list_entry (qtd_list->next, struct ehci_qtd, qtd_list);
+	dev = (struct hcd_dev *)urb->dev->hcpriv;
+	epnum = usb_pipeendpoint (urb->pipe);
+	if (usb_pipein (urb->pipe))
+		epnum |= 0x10;
+
+	vdbg ("%s: submit_async urb %p len %d ep %d-%s qtd %p [qh %p]",
+		ehci->hcd.self.bus_name, urb, urb->transfer_buffer_length,
+		epnum & 0x0f, (epnum & 0x10) ? "in" : "out",
+		qtd, dev ? dev->ep [epnum] : (void *)~0);
+
+	spin_lock_irqsave (&ehci->lock, flags);
+	qh = qh_append_tds (ehci, urb, qtd_list, epnum, &dev->ep [epnum]);
 
 	/* Control/bulk operations through TTs don't need scheduling,
 	 * the HC and TT handle it when the TT has a buffer ready.
 	 */
 	if (likely (qh != 0)) {
-		urb->hcpriv = qh_get (qh);
 		if (likely (qh->qh_state == QH_STATE_IDLE))
 			qh_link_async (ehci, qh_get (qh));
 	}
@@ -873,7 +930,7 @@ static void end_unlink_async (struct ehci_hcd *ehci)
 	ehci->reclaim = 0;
 	ehci->reclaim_ready = 0;
 
-	qh_completions (ehci, qh, 1);
+	qh_completions (ehci, qh);
 
 	// unlink any urb should now unlink all following urbs, so that
 	// relinking only happens for urbs before the unlinked ones.
@@ -973,13 +1030,15 @@ static void scan_async (struct ehci_hcd *ehci)
 				spin_unlock_irqrestore (&ehci->lock, flags);
 
 				/* concurrent unlink could happen here */
-				qh_completions (ehci, qh, 1);
+				qh_completions (ehci, qh);
 
 				spin_lock_irqsave (&ehci->lock, flags);
 				qh_put (ehci, qh);
 			}
 
-			/* unlink idle entries (reduces PCI usage) */
+			/* unlink idle entries, reducing HC PCI usage as
+			 * well as HCD schedule-scanning costs
+			 */
 			if (list_empty (&qh->qtd_list) && !ehci->reclaim) {
 				if (qh->qh_next.qh != qh) {
 					// dbg ("irq/empty");
@@ -987,6 +1046,7 @@ static void scan_async (struct ehci_hcd *ehci)
 				} else {
 					// FIXME:  arrange to stop
 					// after it's been idle a while.
+					// stop/restart isn't free...
 				}
 			}
 			qh = qh->qh_next.qh;

drivers/usb/host/ehci-sched.c

@@ -220,31 +220,31 @@ static int disable_periodic (struct ehci_hcd *ehci)
 
 /*-------------------------------------------------------------------------*/
 
+// FIXME microframe periods not yet handled
+
 static void intr_deschedule (
 	struct ehci_hcd	*ehci,
-	unsigned	frame,
 	struct ehci_qh	*qh,
-	unsigned	period
+	int		wait
 ) {
 	unsigned long	flags;
 	int		status;
-
-	period >>= 3;		// FIXME microframe periods not handled yet
+	unsigned	frame = qh->start;
 
 	spin_lock_irqsave (&ehci->lock, flags);
 
 	do {
 		periodic_unlink (ehci, frame, qh);
 		qh_put (ehci, qh);
-		frame += period;
+		frame += qh->period;
 	} while (frame < ehci->periodic_size);
 
 	qh->qh_state = QH_STATE_UNLINK;
 	qh->qh_next.ptr = 0;
-	ehci->periodic_urbs--;
+	ehci->periodic_sched--;
 
 	/* maybe turn off periodic schedule */
-	if (!ehci->periodic_urbs)
+	if (!ehci->periodic_sched)
 		status = disable_periodic (ehci);
 	else {
 		status = 0;
@@ -258,21 +258,35 @@ static void intr_deschedule (
 	 * (yeech!) to be sure it's done.
 	 * No other threads may be mucking with this qh.
 	 */
-	if (!status && ((ehci_get_frame (&ehci->hcd) - frame) % period) == 0)
-		udelay (125);
+	if (((ehci_get_frame (&ehci->hcd) - frame) % qh->period) == 0) {
+		if (wait) {
+			udelay (125);
+			qh->hw_next = EHCI_LIST_END;
+		} else {
+			/* we may not be IDLE yet, but if the qh is empty
+			 * the race is very short.  then if qh also isn't
+			 * rescheduled soon, it won't matter.  otherwise...
+			 */
+			vdbg ("intr_deschedule...");
+		}
+	} else
+		qh->hw_next = EHCI_LIST_END;
 
 	qh->qh_state = QH_STATE_IDLE;
-	qh->hw_next = EHCI_LIST_END;
+
+	/* update per-qh bandwidth utilization (for usbfs) */
+	ehci->hcd.self.bandwidth_allocated -= 
+		(qh->usecs + qh->c_usecs) / qh->period;
 
 	vdbg ("descheduled qh %p, per = %d frame = %d count = %d, urbs = %d",
-		qh, period, frame,
-		atomic_read (&qh->refcount), ehci->periodic_urbs);
+		qh, qh->period, frame,
+		atomic_read (&qh->refcount), ehci->periodic_sched);
 }
 
 static int check_period (
 	struct ehci_hcd *ehci, 
 	unsigned	frame,
-	int		uframe,
+	unsigned	uframe,
 	unsigned	period,
 	unsigned	usecs
 ) {
@@ -309,19 +323,142 @@ static int check_period (
 	return 1;
 }
 
+static int check_intr_schedule (
+	struct ehci_hcd		*ehci, 
+	unsigned		frame,
+	unsigned		uframe,
+	const struct ehci_qh	*qh,
+	u32			*c_maskp
+)
+{
+    	int		retval = -ENOSPC;
+
+	if (!check_period (ehci, frame, uframe, qh->period, qh->usecs))
+		goto done;
+	if (!qh->c_usecs) {
+		retval = 0;
+		*c_maskp = cpu_to_le32 (0);
+		goto done;
+	}
+
+	/* This is a split transaction; check the bandwidth available for
+	 * the completion too.  Check both worst and best case gaps: worst
+	 * case is SPLIT near uframe end, and CSPLIT near start ... best is
+	 * vice versa.  Difference can be almost two uframe times, but we
+	 * reserve unnecessary bandwidth (waste it) this way.  (Actually
+	 * even better cases exist, like immediate device NAK.)
+	 *
+	 * FIXME don't even bother unless we know this TT is idle in that
+	 * range of uframes ... for now, check_period() allows only one
+	 * interrupt transfer per frame, so needn't check "TT busy" status
+	 * when scheduling a split (QH, SITD, or FSTN).
+	 *
+	 * FIXME ehci 0.96 and above can use FSTNs
+	 */
+	if (!check_period (ehci, frame, uframe + qh->gap_uf + 1,
+				qh->period, qh->c_usecs))
+		goto done;
+	if (!check_period (ehci, frame, uframe + qh->gap_uf,
+				qh->period, qh->c_usecs))
+		goto done;
+
+	*c_maskp = cpu_to_le32 (0x03 << (8 + uframe + qh->gap_uf));
+	retval = 0;
+done:
+	return retval;
+}
+
+static int qh_schedule (struct ehci_hcd *ehci, struct ehci_qh *qh)
+{
+	int 		status;
+	unsigned	uframe;
+	u32		c_mask;
+	unsigned	frame;		/* 0..(qh->period - 1), or NO_FRAME */
+
+	qh->hw_next = EHCI_LIST_END;
+	frame = qh->start;
+
+	/* reuse the previous schedule slots, if we can */
+	if (frame < qh->period) {
+		uframe = ffs (le32_to_cpup (&qh->hw_info2) & 0x00ff);
+		status = check_intr_schedule (ehci, frame, --uframe,
+				qh, &c_mask);
+	} else {
+		uframe = 0;
+		c_mask = 0;
+		status = -ENOSPC;
+	}
+
+	/* else scan the schedule to find a group of slots such that all
+	 * uframes have enough periodic bandwidth available.
+	 */
+	if (status) {
+		frame = qh->period - 1;
+		do {
+			for (uframe = 0; uframe < 8; uframe++) {
+				status = check_intr_schedule (ehci,
+						frame, uframe, qh,
+						&c_mask);
+				if (status == 0)
+					break;
+			}
+		} while (status && --frame);
+		if (status)
+			goto done;
+		qh->start = frame;
+
+		/* reset S-frame and (maybe) C-frame masks */
+		qh->hw_info2 &= ~0xffff;
+		qh->hw_info2 |= cpu_to_le32 (1 << uframe) | c_mask;
+	} else
+		dbg ("reused previous qh %p schedule", qh);
+
+	/* stuff into the periodic schedule */
+	qh->qh_state = QH_STATE_LINKED;
+	dbg ("qh %p usecs %d/%d period %d.0 starting %d.%d (gap %d)",
+		qh, qh->usecs, qh->c_usecs,
+		qh->period, frame, uframe, qh->gap_uf);
+	do {
+		if (unlikely (ehci->pshadow [frame].ptr != 0)) {
+
+// FIXME -- just link toward the end, before any qh with a shorter period,
+// AND accomodate it already having been linked here (after some other qh)
+// AS WELL AS updating the schedule checking logic
+
+			BUG ();
+		} else {
+			ehci->pshadow [frame].qh = qh_get (qh);
+			ehci->periodic [frame] =
+				QH_NEXT (qh->qh_dma);
+		}
+		wmb ();
+		frame += qh->period;
+	} while (frame < ehci->periodic_size);
+
+	/* update per-qh bandwidth for usbfs */
+	ehci->hcd.self.bandwidth_allocated += 
+		(qh->usecs + qh->c_usecs) / qh->period;
+
+	/* maybe enable periodic schedule processing */
+	if (!ehci->periodic_sched++)
+		status = enable_periodic (ehci);
+done:
+	return status;
+}
+
 static int intr_submit (
 	struct ehci_hcd		*ehci,
 	struct urb		*urb,
 	struct list_head	*qtd_list,
 	int			mem_flags
 ) {
-	unsigned		epnum, period;
-	unsigned short		usecs, c_usecs, gap_uf;
+	unsigned		epnum;
 	unsigned long		flags;
 	struct ehci_qh		*qh;
 	struct hcd_dev		*dev;
 	int			is_input;
 	int			status = 0;
+	struct list_head	empty;
 
 	/* get endpoint and transfer/schedule data */
 	epnum = usb_pipeendpoint (urb->pipe);
@@ -329,198 +466,30 @@ static int intr_submit (
 	if (is_input)
 		epnum |= 0x10;
 
-	/*
-	 * HS interrupt transfers are simple -- only one microframe.  FS/LS
-	 * interrupt transfers involve a SPLIT in one microframe and CSPLIT
-	 * sometime later.  We need to know how much time each will be
-	 * needed in each microframe and, for FS/LS, how many microframes
-	 * separate the two in the best case.
-	 */
-	usecs = usb_calc_bus_time (USB_SPEED_HIGH, is_input, 0,
-			urb->transfer_buffer_length);
-	if (urb->dev->speed == USB_SPEED_HIGH) {
-		gap_uf = 0;
-		c_usecs = 0;
-
-		/* FIXME handle HS periods of less than 1 frame. */
-		period = urb->interval >> 3;
-		if (period < 1) {
-			dbg ("intr period %d uframes, NYET!", urb->interval);
-			status = -EINVAL;
-			goto done;
-		}
-	} else {
-		/* gap is a function of full/low speed transfer times */
-		gap_uf = 1 + usb_calc_bus_time (urb->dev->speed, is_input, 0,
-				urb->transfer_buffer_length) / (125 * 1000);
-
-		/* FIXME this just approximates SPLIT/CSPLIT times */
-		if (is_input) {		// SPLIT, gap, CSPLIT+DATA
-			c_usecs = usecs + HS_USECS (0);
-			usecs = HS_USECS (1);
-		} else {		// SPLIT+DATA, gap, CSPLIT
-			usecs = usecs + HS_USECS (1);
-			c_usecs = HS_USECS (0);
-		}
-
-		period = urb->interval;
-	}
+	spin_lock_irqsave (&ehci->lock, flags);
+	dev = (struct hcd_dev *)urb->dev->hcpriv;
 
-	/*
-	 * NOTE: current completion/restart logic doesn't handle more than
-	 * one qtd in a periodic qh ... 16-20 KB/urb is pretty big for this.
-	 * such big requests need many periods to transfer.
-	 *
-	 * FIXME want to change hcd core submit model to expect queuing
-	 * for all transfer types ... not just ISO and (with flag) BULK.
-	 * that means: getting rid of this check; handling the "interrupt
-	 * urb already queued" case below like bulk queuing is handled (no
-	 * errors possible!); and completly getting rid of that annoying
-	 * qh restart logic.  simpler/smaller overall, and more flexible.
-	 */
-	if (unlikely (qtd_list->next != qtd_list->prev)) {
-		dbg ("only one intr qtd per urb allowed"); 
-		status = -EINVAL;
+	/* get qh and force any scheduling errors */
+	INIT_LIST_HEAD (&empty);
+	qh = qh_append_tds (ehci, urb, &empty, epnum, &dev->ep [epnum]);
+	if (qh == 0) {
+		status = -ENOMEM;
 		goto done;
 	}
-
-	spin_lock_irqsave (&ehci->lock, flags);
-
-	/* get the qh (must be empty and idle) */
-	dev = (struct hcd_dev *)urb->dev->hcpriv;
-	qh = (struct ehci_qh *) dev->ep [epnum];
-	if (qh) {
-		/* only allow one queued interrupt urb per EP */
-		if (unlikely (qh->qh_state != QH_STATE_IDLE
-				|| !list_empty (&qh->qtd_list))) {
-			dbg ("interrupt urb already queued");
-			status = -EBUSY;
-		} else {
-			/* maybe reset hardware's data toggle in the qh */
-			if (unlikely (!usb_gettoggle (urb->dev, epnum & 0x0f,
-					!(epnum & 0x10)))) {
-				qh->hw_token |=
-					__constant_cpu_to_le32 (QTD_TOGGLE);
-				usb_settoggle (urb->dev, epnum & 0x0f,
-					!(epnum & 0x10), 1);
-			}
-			/* trust the QH was set up as interrupt ... */
-			list_splice (qtd_list, &qh->qtd_list);
-			qh_update (qh, list_entry (qtd_list->next,
-						struct ehci_qtd, qtd_list));
-			qtd_list = &qh->qtd_list;
-		}
-	} else {
-		/* can't sleep here, we have ehci->lock... */
-		qh = ehci_qh_make (ehci, urb, qtd_list, SLAB_ATOMIC);
-		if (likely (qh != 0)) {
-			// dbg ("new INTR qh %p", qh);
-			dev->ep [epnum] = qh;
-			qtd_list = &qh->qtd_list;
-		} else
-			status = -ENOMEM;
+	if (qh->qh_state == QH_STATE_IDLE) {
+		if ((status = qh_schedule (ehci, qh)) != 0)
+			goto done;
 	}
 
-	/* Schedule this periodic QH. */
-	if (likely (status == 0)) {
-		unsigned	frame = period;
-
-		qh->hw_next = EHCI_LIST_END;
-		qh->usecs = usecs;
-		qh->c_usecs = c_usecs;
-
-		urb->hcpriv = qh_get (qh);
-		status = -ENOSPC;
-
-		/* pick a set of schedule slots, link the QH into them */
-		do {
-			int	uframe;
-			u32	c_mask = 0;
-
-			/* pick a set of slots such that all uframes have
-			 * enough periodic bandwidth available.
-			 */
-			frame--;
-			for (uframe = 0; uframe < 8; uframe++) {
-				if (check_period (ehci, frame, uframe,
-						period, usecs) == 0)
-					continue;
-
-				/* If this is a split transaction, check the
-				 * bandwidth available for the completion
-				 * too.  check both best and worst case gaps:
-				 * worst case is SPLIT near uframe end, and
-				 * CSPLIT near start ... best is vice versa.
-				 * Difference can be almost two uframe times.
-				 *
-				 * FIXME don't even bother unless we know
-				 * this TT is idle in that uframe ... right
-				 * now we know only one interrupt transfer
-				 * will be scheduled per frame, so we don't
-				 * need to update/check TT state when we
-				 * schedule a split (QH, SITD, or FSTN).
-				 *
-				 * FIXME ehci 0.96 and above can use FSTNs
-				 */
-				if (!c_usecs)
-				    	break;
-				if (check_period (ehci, frame,
-						uframe + gap_uf,
-						period, c_usecs) == 0)
-					continue;
-				if (check_period (ehci, frame,
-						uframe + gap_uf + 1,
-						period, c_usecs) == 0)
-					continue;
-
-				c_mask = 0x03 << (8 + uframe + gap_uf);
-				c_mask = cpu_to_le32 (c_mask);
-				break;
-			}
-			if (uframe == 8)
-				continue;
-
-			/* QH will run once each period, starting there  */
-			urb->start_frame = frame;
-			status = 0;
-
-			/* reset S-frame and (maybe) C-frame masks */
-			qh->hw_info2 &= ~0xffff;
-			qh->hw_info2 |= cpu_to_le32 (1 << uframe) | c_mask;
-			// dbg_qh ("Schedule INTR qh", ehci, qh);
-
-			/* stuff into the periodic schedule */
-			qh->qh_state = QH_STATE_LINKED;
-			vdbg ("qh %p usecs %d period %d.0 starting %d.%d",
-				qh, qh->usecs, period, frame, uframe);
-			do {
-				if (unlikely (ehci->pshadow [frame].ptr != 0)) {
-// FIXME -- just link toward the end, before any qh with a shorter period,
-// AND handle it already being (implicitly) linked into this frame
-// AS WELL AS updating the check_period() logic
-					BUG ();
-				} else {
-					ehci->pshadow [frame].qh = qh_get (qh);
-					ehci->periodic [frame] =
-						QH_NEXT (qh->qh_dma);
-				}
-				wmb ();
-				frame += period;
-			} while (frame < ehci->periodic_size);
-
-			/* update bandwidth utilization records (for usbfs) */
-			usb_claim_bandwidth (urb->dev, urb,
-				(usecs + c_usecs) / period, 0);
+	/* then queue the urb's tds to the qh */
+	qh = qh_append_tds (ehci, urb, qtd_list, epnum, &dev->ep [epnum]);
+	BUG_ON (qh == 0);
 
-			/* maybe enable periodic schedule processing */
-			if (!ehci->periodic_urbs++)
-				status = enable_periodic (ehci);
-			break;
+	/* ... update usbfs periodic stats */
+	ehci->hcd.self.bandwidth_int_reqs++;
 
-		} while (frame);
-	}
-	spin_unlock_irqrestore (&ehci->lock, flags);
 done:
+	spin_unlock_irqrestore (&ehci->lock, flags);
 	if (status)
 		qtd_list_free (ehci, urb, qtd_list);
 
@@ -534,10 +503,6 @@ intr_complete (
 	struct ehci_qh	*qh,
 	unsigned long	flags		/* caller owns ehci->lock ... */
 ) {
-	struct ehci_qtd	*qtd;
-	struct urb	*urb;
-	int		unlinking;
-
 	/* nothing to report? */
 	if (likely ((qh->hw_token & __constant_cpu_to_le32 (QTD_STS_ACTIVE))
 			!= 0))
@@ -547,43 +512,14 @@ intr_complete (
 		return flags;
 	}
 	
-	qtd = list_entry (qh->qtd_list.next, struct ehci_qtd, qtd_list);
-	urb = qtd->urb;
-	unlinking = (urb->status == -ENOENT) || (urb->status == -ECONNRESET);
-
-	/* call any completions, after patching for reactivation */
+	/* handle any completions */
 	spin_unlock_irqrestore (&ehci->lock, flags);
-	/* NOTE:  currently restricted to one qtd per qh! */
-	if (qh_completions (ehci, qh, 0) == 0)
-		urb = 0;
+	qh_completions (ehci, qh);
 	spin_lock_irqsave (&ehci->lock, flags);
 
-	/* never reactivate requests that were unlinked ... */
-	if (likely (urb != 0)) {
-		if (unlinking
-				|| urb->status == -ECONNRESET
-				|| urb->status == -ENOENT
-				// || (urb->dev == null)
-				|| ehci->hcd.state == USB_STATE_HALT)
-			urb = 0;
-		// FIXME look at all those unlink cases ... we always
-		// need exactly one completion that reports unlink.
-		// the one above might not have been it!
-	}
+	if (unlikely (list_empty (&qh->qtd_list)))
+		intr_deschedule (ehci, qh, 0);
 
-	/* normally reactivate */
-	if (likely (urb != 0)) {
-		if (usb_pipeout (urb->pipe))
-			pci_dma_sync_single (ehci->hcd.pdev,
-				qtd->buf_dma,
-				urb->transfer_buffer_length,
-				PCI_DMA_TODEVICE);
-		urb->status = -EINPROGRESS;
-		urb->actual_length = 0;
-
-		/* patch qh and restart */
-		qh_update (qh, qtd);
-	}
 	return flags;
 }
 
@@ -806,7 +742,7 @@ static int get_iso_range (
 
 	/* calculate the legal range [start,max) */
 	now = readl (&ehci->regs->frame_index) + 1;	/* next uframe */
-	if (!ehci->periodic_urbs)
+	if (!ehci->periodic_sched)
 		now += 8;				/* startup delay */
 	now %= mod;
 	end = now + mod;
@@ -926,7 +862,7 @@ itd_schedule (struct ehci_hcd *ehci, struct urb *urb)
 		usb_claim_bandwidth (urb->dev, urb, usecs, 1);
 
 		/* maybe enable periodic schedule processing */
-		if (!ehci->periodic_urbs++) {
+		if (!ehci->periodic_sched++) {
 			if ((status =  enable_periodic (ehci)) != 0) {
 				// FIXME deschedule right away
 				err ("itd_schedule, enable = %d", status);
@@ -1009,8 +945,8 @@ itd_complete (
 	spin_lock_irqsave (&ehci->lock, flags);
 
 	/* defer stopping schedule; completion can submit */
-	ehci->periodic_urbs--;
-	if (!ehci->periodic_urbs)
+	ehci->periodic_sched--;
+	if (!ehci->periodic_sched)
 		(void) disable_periodic (ehci);
 
 	return flags;

drivers/usb/host/ehci.h

@@ -50,7 +50,7 @@ struct ehci_hcd {			/* one per controller */
 
 	union ehci_shadow	*pshadow;	/* mirror hw periodic table */
 	int			next_uframe;	/* scan periodic, start here */
-	unsigned		periodic_urbs;	/* how many urbs scheduled? */
+	unsigned		periodic_sched;	/* periodic activity count */
 
 	/* deferred work from IRQ, etc */
 	struct tasklet_struct	tasklet;
@@ -72,7 +72,7 @@ struct ehci_hcd {			/* one per controller */
 };
 
 /* unwrap an HCD pointer to get an EHCI_HCD pointer */ 
-#define hcd_to_ehci(hcd_ptr) list_entry(hcd_ptr, struct ehci_hcd, hcd)
+#define hcd_to_ehci(hcd_ptr) container_of(hcd_ptr, struct ehci_hcd, hcd)
 
 /* NOTE:  urb->transfer_flags expected to not use this bit !!! */
 #define EHCI_STATE_UNLINK	0x8000		/* urb being unlinked */
@@ -287,12 +287,20 @@ struct ehci_qh {
 	struct list_head	qtd_list;	/* sw qtd list */
 
 	atomic_t		refcount;
-	unsigned short		usecs;		/* intr bandwidth */
-	unsigned short		c_usecs;	/* ... split completion bw */
-	short			qh_state;
+
+	u8			qh_state;
 #define	QH_STATE_LINKED		1		/* HC sees this */
 #define	QH_STATE_UNLINK		2		/* HC may still see this */
 #define	QH_STATE_IDLE		3		/* HC doesn't see this */
+
+	/* periodic schedule info */
+	u8			usecs;		/* intr bandwidth */
+	u8			gap_uf;		/* uframes split/csplit gap */
+	u8			c_usecs;	/* ... split completion bw */
+	unsigned short		period;		/* polling interval */
+	unsigned short		start;		/* where polling starts */
+#define NO_FRAME ((unsigned short)~0)			/* pick new start */
+
 } __attribute__ ((aligned (32)));
 
 /*-------------------------------------------------------------------------*/

drivers/usb/host/ohci-q.c

@@ -938,7 +938,7 @@ static void finish_unlinks (struct ohci_hcd *ohci, u16 tick)
 		/* ED's now officially unlinked, hc doesn't see */
 		ed->state = ED_IDLE;
 		ed->hwINFO &= ~ED_SKIP;
-		ed->hwHeadP &= ~cpu_to_le32 (ED_H);
+		ed->hwHeadP &= ~ED_H;
 		ed->hwNextED = 0;
 
 		/* but if there's work queued, reschedule */

drivers/usb/media/konicawc.c

 /*
- * $Id$
- *
  * konicawc.c - konica webcam driver
  *
  * Author: Simon Evans <spse@secret.org.uk>
@@ -8,7 +6,7 @@
  * Copyright (C) 2002 Simon Evans
  *
  * Licence: GPL
- * 
+ *
  * Driver for USB webcams based on Konica chipset. This
  * chipset is used in Intel YC76 camera.
  *
@@ -18,6 +16,7 @@
 #include <linux/module.h>
 #include <linux/init.h>
 
+
 #include "usbvideo.h"
 
 #define MAX_BRIGHTNESS	108
@@ -26,9 +25,11 @@
 #define MAX_SHARPNESS	108
 #define MAX_WHITEBAL	372
 #define MAX_SPEED	6
+
+
 #define MAX_CAMERAS	1
 
-#define DRIVER_VERSION	"v1.1"
+#define DRIVER_VERSION	"v1.3"
 #define DRIVER_DESC	"Konica Webcam driver"
 
 enum ctrl_req {
@@ -41,18 +42,32 @@ enum ctrl_req {
 
 
 enum frame_sizes {
-	SIZE_160X136	= 0,
-	SIZE_176X144	= 1,
-	SIZE_320X240	= 2,
+	SIZE_160X120	= 0,
+	SIZE_160X136	= 1,
+	SIZE_176X144	= 2,
+	SIZE_320X240	= 3,
+	
 };
 
+#define MAX_FRAME_SIZE	SIZE_320X240
 
 static usbvideo_t *cams;
 
+#ifdef CONFIG_USB_DEBUG
+static int debug;
+#define DEBUG(n, format, arg...) \
+	if (n <= debug) {	 \
+		printk(KERN_DEBUG __FILE__ ":%s(): " format "\n", __FUNCTION__ , ## arg); \
+	}
+#else
+#define DEBUG(n, arg...)
+static const int debug = 0;
+#endif
+
+
 /* Some default values for inital camera settings,
    can be set by modprobe */
 
-static int debug;
 static enum frame_sizes size;	
 static int speed = 6;		/* Speed (fps) 0 (slowest) to 6 (fastest) */
 static int brightness =	MAX_BRIGHTNESS/2;
@@ -61,31 +76,36 @@ static int saturation =	MAX_SATURATION/2;
 static int sharpness =	MAX_SHARPNESS/2;
 static int whitebal =	3*(MAX_WHITEBAL/4);
 
-static int speed_to_interface[] = { 1, 0, 3, 2, 4, 5, 6 };
+static int spd_to_iface[] = { 1, 0, 3, 2, 4, 5, 6 };
 
 /* These FPS speeds are from the windows config box. They are
  * indexed on size (0-2) and speed (0-6). Divide by 3 to get the
  * real fps.
  */
 
-static int speed_to_fps[3][7] = { { 24, 40, 48, 60, 72, 80, 100 },
-				  { 18, 30, 36, 45, 54, 60, 75  },
-				  { 6,  10, 12, 15, 18, 20, 25  } };
-
+static int spd_to_fps[][7] = { { 24, 40, 48, 60, 72, 80, 100 },
+			       { 24, 40, 48, 60, 72, 80, 100 },
+			       { 18, 30, 36, 45, 54, 60, 75  },
+			       { 6,  10, 12, 15, 18, 21, 25  } };
 
-static int camera_sizes[][2] = { { 160, 136 },
-				 { 176, 144 },
-				 { 320, 240 },
-				 { } /* List terminator */
+struct cam_size {
+	u16	width;
+	u16	height;
+	u8	cmd;
 };
 
+static struct cam_size camera_sizes[] = { { 160, 120, 0x7 },
+					  { 160, 136, 0xa },
+					  { 176, 144, 0x4 },
+					  { 320, 240, 0x5 } };
+
 struct konicawc {
 	u8 brightness;		/* camera uses 0 - 9, x11 for real value */
 	u8 contrast;		/* as above */
 	u8 saturation;		/* as above */
 	u8 sharpness;		/* as above */
 	u8 white_bal;		/* 0 - 33, x11 for real value */
-	u8 speed;		/* Stored as 0 - 6, used as index in speed_to_* (above) */
+	u8 speed;		/* Stored as 0 - 6, used as index in spd_to_* (above) */
 	u8 size;		/* Frame Size */
 	int height;
 	int width;
@@ -93,6 +113,10 @@ struct konicawc {
 	u8 sts_buf[USBVIDEO_NUMSBUF][FRAMES_PER_DESC];
 	struct urb *last_data_urb;
 	int lastframe;
+	int cur_frame_size;	/* number of bytes in current frame size */
+	int maxline;		/* number of lines per frame */
+	int yplanesz;		/* Number of bytes in the Y plane */
+	unsigned int buttonsts:1;
 };
 
 
@@ -110,42 +134,56 @@ static int konicawc_ctrl_msg(uvd_t *uvd, u8 dir, u8 request, u16 value, u16 inde
 }
 
 
+static inline void konicawc_camera_on(uvd_t *uvd)
+{
+        DEBUG(0, "camera on");
+        konicawc_set_misc(uvd, 0x2, 1, 0x0b);
+}
+
+
+static inline void konicawc_camera_off(uvd_t *uvd)
+{
+        DEBUG(0, "camera off");
+        konicawc_set_misc(uvd, 0x2, 0, 0x0b);
+}
+
+
+static void konicawc_set_camera_size(uvd_t *uvd)
+{
+	struct konicawc *cam = (struct konicawc *)uvd->user_data;
+
+	konicawc_set_misc(uvd, 0x2, camera_sizes[cam->size].cmd, 0x08);
+	cam->width = camera_sizes[cam->size].width;
+	cam->height = camera_sizes[cam->size].height;
+	cam->yplanesz = cam->height * cam->width;
+	cam->cur_frame_size = (cam->yplanesz * 3) / 2;
+	cam->maxline = cam->yplanesz / 256;
+	uvd->videosize = VIDEOSIZE(cam->width, cam->height);
+}
+
+
 static int konicawc_setup_on_open(uvd_t *uvd)
 {
 	struct konicawc *cam = (struct konicawc *)uvd->user_data;
 
-	konicawc_set_misc(uvd, 0x2, 0, 0x0b);
-	dbg("setting brightness to %d (%d)", cam->brightness,
+	DEBUG(1, "setting brightness to %d (%d)", cam->brightness,
 	    cam->brightness * 11);
 	konicawc_set_value(uvd, cam->brightness, SetBrightness);
-	dbg("setting white balance to %d (%d)", cam->white_bal,
+	DEBUG(1, "setting white balance to %d (%d)", cam->white_bal,
 	    cam->white_bal * 11);
 	konicawc_set_value(uvd, cam->white_bal, SetWhitebal);
-	dbg("setting contrast to %d (%d)", cam->contrast,
+	DEBUG(1, "setting contrast to %d (%d)", cam->contrast,
 	    cam->contrast * 11);
 	konicawc_set_value(uvd, cam->contrast, SetContrast);
-	dbg("setting saturation to %d (%d)", cam->saturation,
+	DEBUG(1, "setting saturation to %d (%d)", cam->saturation,
 	    cam->saturation * 11);
 	konicawc_set_value(uvd, cam->saturation, SetSaturation);
-	dbg("setting sharpness to %d (%d)", cam->sharpness,
+	DEBUG(1, "setting sharpness to %d (%d)", cam->sharpness,
 	    cam->sharpness * 11);
 	konicawc_set_value(uvd, cam->sharpness, SetSharpness);
-	dbg("setting size %d", cam->size);
-	switch(cam->size) {
-	case 0:
-		konicawc_set_misc(uvd, 0x2, 0xa, 0x08);
-		break;
-
-	case 1:
-		konicawc_set_misc(uvd, 0x2, 4, 0x08);
-		break;
-
-	case 2:
-		konicawc_set_misc(uvd, 0x2, 5, 0x08);
-		break;
-	}
-	konicawc_set_misc(uvd, 0x2, 1, 0x0b);
-	cam->lastframe = -1;
+	konicawc_set_camera_size(uvd);
+	cam->lastframe = -2;
+	cam->buttonsts = 0;
 	return 0;
 }
 
@@ -154,23 +192,25 @@ static void konicawc_adjust_picture(uvd_t *uvd)
 {
 	struct konicawc *cam = (struct konicawc *)uvd->user_data;
 
-	dbg("new brightness: %d", uvd->vpic.brightness);
+	konicawc_camera_off(uvd);
+	DEBUG(1, "new brightness: %d", uvd->vpic.brightness);
 	uvd->vpic.brightness = (uvd->vpic.brightness > MAX_BRIGHTNESS) ? MAX_BRIGHTNESS : uvd->vpic.brightness;
 	if(cam->brightness != uvd->vpic.brightness / 11) {
 	   cam->brightness = uvd->vpic.brightness / 11;
-	   dbg("setting brightness to %d (%d)", cam->brightness,
+	   DEBUG(1, "setting brightness to %d (%d)", cam->brightness,
 	       cam->brightness * 11);
 	   konicawc_set_value(uvd, cam->brightness, SetBrightness);
 	}
 
-	dbg("new contrast: %d", uvd->vpic.contrast);
+	DEBUG(1, "new contrast: %d", uvd->vpic.contrast);
 	uvd->vpic.contrast = (uvd->vpic.contrast > MAX_CONTRAST) ? MAX_CONTRAST : uvd->vpic.contrast;
 	if(cam->contrast != uvd->vpic.contrast / 11) {
 		cam->contrast = uvd->vpic.contrast / 11;
-		dbg("setting contrast to %d (%d)", cam->contrast,
+		DEBUG(1, "setting contrast to %d (%d)", cam->contrast,
 		    cam->contrast * 11);
 		konicawc_set_value(uvd, cam->contrast, SetContrast);
 	}
+	konicawc_camera_on(uvd);
 }
 
 
@@ -180,21 +220,20 @@ static int konicawc_compress_iso(uvd_t *uvd, struct urb *dataurb, struct urb *st
 	int i, totlen = 0;
 	unsigned char *status = stsurb->transfer_buffer;
 	int keep = 0, discard = 0, bad = 0;
-	static int buttonsts = 0;
+	struct konicawc *cam = (struct konicawc *)uvd->user_data;
 
 	for (i = 0; i < dataurb->number_of_packets; i++) {
-		int button = buttonsts;
+		int button = cam->buttonsts;
 		unsigned char sts;
 		int n = dataurb->iso_frame_desc[i].actual_length;
 		int st = dataurb->iso_frame_desc[i].status;
-		cdata = dataurb->transfer_buffer + 
+		cdata = dataurb->transfer_buffer +
 			dataurb->iso_frame_desc[i].offset;
 
 		/* Detect and ignore errored packets */
 		if (st < 0) {
-			if (debug >= 1)
-				err("Data error: packet=%d. len=%d. status=%d.",
-				    i, n, st);
+			DEBUG(1, "Data error: packet=%d. len=%d. status=%d.",
+			      i, n, st);
 			uvd->stats.iso_err_count++;
 			continue;
 		}
@@ -210,8 +249,8 @@ static int konicawc_compress_iso(uvd_t *uvd, struct urb *dataurb, struct urb *st
 
 		/* sts: 0x80-0xff: frame start with frame number (ie 0-7f)
 		 * otherwise:
-		 * bit 0 0:drop packet (padding data)
-		 *	 1 keep packet
+		 * bit 0 0: keep packet
+		 *	 1: drop packet (padding data)
 		 *
 		 * bit 4 0 button not clicked
 		 *       1 button clicked
@@ -225,10 +264,10 @@ static int konicawc_compress_iso(uvd_t *uvd, struct urb *dataurb, struct urb *st
 		
 		/* work out the button status, but dont do
 		   anything with it for now */
-		   
-		if(button != buttonsts) {
-			dbg("button: %sclicked", button ? "" : "un");
-			buttonsts = button;
+
+		if(button != cam->buttonsts) {
+			DEBUG(2, "button: %sclicked", button ? "" : "un");
+			cam->buttonsts = button;
 		}
 
 		if(sts == 0x01) { /* drop frame */
@@ -241,34 +280,52 @@ static int konicawc_compress_iso(uvd_t *uvd, struct urb *dataurb, struct urb *st
 			bad++;
 			continue;
 		}
+		if(!sts && cam->lastframe == -2) {
+			DEBUG(2, "dropping frame looking for image start");
+			continue;
+		}
 
 		keep++;
-		if(*(status+i) & 0x80) { /* frame start */
+		if(sts & 0x80) { /* frame start */
 			unsigned char marker[] = { 0, 0xff, 0, 0x00 };
-			if(debug > 1)
-				dbg("Adding Marker packet = %d, frame = %2.2x",
-				    i, *(status+i));
-			marker[3] = *(status+i) - 0x80;
-			RingQueue_Enqueue(&uvd->dp, marker, 4);			
+
+			if(cam->lastframe == -2) {
+				DEBUG(2, "found initial image");
+				cam->lastframe = -1;
+			}
+				
+			marker[3] = sts & 0x7F;
+			RingQueue_Enqueue(&uvd->dp, marker, 4);
 			totlen += 4;
 		}
+
 		totlen += n;	/* Little local accounting */
-		if(debug > 5)
-			dbg("Adding packet %d, bytes = %d", i, n);
 		RingQueue_Enqueue(&uvd->dp, cdata, n);
-
 	}
-	if(debug > 8) {
-		dbg("finished: keep = %d discard = %d bad = %d added %d bytes",
+	DEBUG(8, "finished: keep = %d discard = %d bad = %d added %d bytes",
 		    keep, discard, bad, totlen);
-	}
 	return totlen;
 }
 
 
+static void resubmit_urb(uvd_t *uvd, struct urb *urb)
+{
+        int i, ret;
+        for (i = 0; i < FRAMES_PER_DESC; i++) {
+                urb->iso_frame_desc[i].status = 0;
+        }
+        urb->dev = uvd->dev;
+        urb->status = 0;
+        ret = usb_submit_urb(urb, GFP_KERNEL);
+	DEBUG(3, "submitting urb of length %d", urb->transfer_buffer_length);
+        if(ret)
+                err("usb_submit_urb error (%d)", ret);
+
+}
+
+
 static void konicawc_isoc_irq(struct urb *urb)
 {
-	int i, ret, len = 0;
 	uvd_t *uvd = urb->context;
 	struct konicawc *cam = (struct konicawc *)uvd->user_data;
 
@@ -277,42 +334,35 @@ static void konicawc_isoc_irq(struct urb *urb)
 		return;
 
 	if (!uvd->streaming) {
-		if (debug >= 1)
-			info("Not streaming, but interrupt!");
+		DEBUG(1, "Not streaming, but interrupt!");
 		return;
 	}
 
-	if (urb->actual_length > 32) {
-		cam->last_data_urb = urb;
-		goto urb_done_with;
-	}
+	DEBUG(3, "got frame %d len = %d buflen =%d", urb->start_frame, urb->actual_length, urb->transfer_buffer_length);
 
 	uvd->stats.urb_count++;
-	if (urb->actual_length <= 0)
-		goto urb_done_with;
 
+	if (urb->transfer_buffer_length > 32) {
+		cam->last_data_urb = urb;
+		return;
+	}
 	/* Copy the data received into ring queue */
 	if(cam->last_data_urb) {
-		len = konicawc_compress_iso(uvd, cam->last_data_urb, urb);
-		for (i = 0; i < FRAMES_PER_DESC; i++) {
-			cam->last_data_urb->iso_frame_desc[i].status = 0;
-		}
+		int len = 0;
+		if(urb->start_frame != cam->last_data_urb->start_frame)
+			err("Lost sync on frames");
+		else if (!urb->status && !cam->last_data_urb->status)
+			len = konicawc_compress_iso(uvd, cam->last_data_urb, urb);
+
+		resubmit_urb(uvd, urb);
+		resubmit_urb(uvd, cam->last_data_urb);
 		cam->last_data_urb = NULL;
+		uvd->stats.urb_length = len;
+		uvd->stats.data_count += len;
+		if(len)
+			RingQueue_WakeUpInterruptible(&uvd->dp);
+		return;
 	}
-	uvd->stats.urb_length = len;
-	uvd->stats.data_count += len;
-	if(len)
-		RingQueue_WakeUpInterruptible(&uvd->dp);
-
-urb_done_with:
-	for (i = 0; i < FRAMES_PER_DESC; i++) {
-		urb->iso_frame_desc[i].status = 0;
-	}
-	urb->dev = uvd->dev;
-	urb->status = 0;
-	ret = usb_submit_urb(urb, GFP_KERNEL);
-	if(ret)
-		err("usb_submit_urb error (%d)", ret);
 	return;
 }
 
@@ -322,13 +372,18 @@ static int konicawc_start_data(uvd_t *uvd)
 	struct usb_device *dev = uvd->dev;
 	int i, errFlag;
 	struct konicawc *cam = (struct konicawc *)uvd->user_data;
+	int pktsz;
+	struct usb_interface_descriptor *interface;
 
+	interface = &dev->actconfig->interface[uvd->iface].altsetting[spd_to_iface[cam->speed]];
+	pktsz = interface->endpoint[1].wMaxPacketSize;
+	DEBUG(1, "pktsz = %d", pktsz);
 	if (!CAMERA_IS_OPERATIONAL(uvd)) {
 		err("Camera is not operational");
 		return -EFAULT;
 	}
 	uvd->curframe = -1;
-
+	konicawc_camera_on(uvd);
 	/* Alternate interface 1 is is the biggest frame size */
 	i = usb_set_interface(dev, uvd->iface, uvd->ifaceAltActive);
 	if (i < 0) {
@@ -349,10 +404,10 @@ static int konicawc_start_data(uvd_t *uvd)
 		urb->transfer_buffer = uvd->sbuf[i].data;
 		urb->complete = konicawc_isoc_irq;
 		urb->number_of_packets = FRAMES_PER_DESC;
-		urb->transfer_buffer_length = uvd->iso_packet_len * FRAMES_PER_DESC;
-		for (j=k=0; j < FRAMES_PER_DESC; j++, k += uvd->iso_packet_len) {
+		urb->transfer_buffer_length = pktsz * FRAMES_PER_DESC;
+		for (j=k=0; j < FRAMES_PER_DESC; j++, k += pktsz) {
 			urb->iso_frame_desc[j].offset = k;
-			urb->iso_frame_desc[j].length = uvd->iso_packet_len;
+			urb->iso_frame_desc[j].length = pktsz;
 		}
 
 		urb = cam->sts_urb[i];
@@ -375,18 +430,17 @@ static int konicawc_start_data(uvd_t *uvd)
 	
 	/* Submit all URBs */
 	for (i=0; i < USBVIDEO_NUMSBUF; i++) {
-		errFlag = usb_submit_urb(uvd->sbuf[i].urb, GFP_KERNEL);
-		if (errFlag)
-			err ("usb_submit_isoc(%d) ret %d", i, errFlag);
-
 		errFlag = usb_submit_urb(cam->sts_urb[i], GFP_KERNEL);
 		if (errFlag)
 			err("usb_submit_isoc(%d) ret %d", i, errFlag);
+
+		errFlag = usb_submit_urb(uvd->sbuf[i].urb, GFP_KERNEL);
+		if (errFlag)
+			err ("usb_submit_isoc(%d) ret %d", i, errFlag);
 	}
 
 	uvd->streaming = 1;
-	if (debug > 1)
-		dbg("streaming=1 video_endp=$%02x", uvd->video_endp);
+	DEBUG(1, "streaming=1 video_endp=$%02x", uvd->video_endp);
 	return 0;
 }
 
@@ -399,6 +453,8 @@ static void konicawc_stop_data(uvd_t *uvd)
 	if ((uvd == NULL) || (!uvd->streaming) || (uvd->dev == NULL))
 		return;
 
+	konicawc_camera_off(uvd);
+	uvd->streaming = 0;
 	cam = (struct konicawc *)uvd->user_data;
 	cam->last_data_urb = NULL;
 
@@ -413,8 +469,6 @@ static void konicawc_stop_data(uvd_t *uvd)
 			err("usb_unlink_urb() error %d.", j);
 	}
 
-	uvd->streaming = 0;
-
 	if (!uvd->remove_pending) {
 		/* Set packet size to 0 */
 		j = usb_set_interface(uvd->dev, uvd->iface, uvd->ifaceAltInactive);
@@ -428,42 +482,33 @@ static void konicawc_stop_data(uvd_t *uvd)
 
 static void konicawc_process_isoc(uvd_t *uvd, usbvideo_frame_t *frame)
 {	
-	int n;
-	int maxline, yplanesz;
 	struct konicawc *cam = (struct konicawc *)uvd->user_data;
-	assert(uvd != NULL);
+	int maxline = cam->maxline;
+	int yplanesz = cam->yplanesz;
+
 	assert(frame != NULL);
 
-	maxline = (cam->height * cam->width * 3) / (2 * 384);
-	yplanesz = cam->height * cam->width;
-	if(debug > 5)
-		dbg("maxline = %d yplanesz = %d", maxline, yplanesz);
-	
-	if(debug > 3)
-		dbg("Frame state = %d", frame->scanstate);
+	DEBUG(5, "maxline = %d yplanesz = %d", maxline, yplanesz);
+	DEBUG(3, "Frame state = %d", frame->scanstate);
 
 	if(frame->scanstate == ScanState_Scanning) {
 		int drop = 0;
 		int curframe;
 		int fdrops = 0;
-		if(debug > 3)
-			dbg("Searching for marker, queue len = %d", RingQueue_GetLength(&uvd->dp));
+		DEBUG(3, "Searching for marker, queue len = %d", RingQueue_GetLength(&uvd->dp));
 		while(RingQueue_GetLength(&uvd->dp) >= 4) {
 			if ((RING_QUEUE_PEEK(&uvd->dp, 0) == 0x00) &&
 			    (RING_QUEUE_PEEK(&uvd->dp, 1) == 0xff) &&
 			    (RING_QUEUE_PEEK(&uvd->dp, 2) == 0x00) &&
 			    (RING_QUEUE_PEEK(&uvd->dp, 3) < 0x80)) {
 				curframe = RING_QUEUE_PEEK(&uvd->dp, 3);
-				if(cam->lastframe != -1) {
-					if(curframe < cam->lastframe) {
-						fdrops = (curframe + 0x80) - cam->lastframe;
-					} else {
-						fdrops = curframe - cam->lastframe;
-					}
+				if(cam->lastframe >= 0) {
+					fdrops = (0x80 + curframe - cam->lastframe) & 0x7F;
 					fdrops--;
-					if(fdrops)
+					if(fdrops) {
 						info("Dropped %d frames (%d -> %d)", fdrops,
 						     cam->lastframe, curframe);
+					}
 				}
 				cam->lastframe = curframe;
 				frame->curline = 0;
@@ -474,18 +519,20 @@ static void konicawc_process_isoc(uvd_t *uvd, usbvideo_frame_t *frame)
 			RING_QUEUE_DEQUEUE_BYTES(&uvd->dp, 1);
 			drop++;
 		}
+		if(drop)
+			DEBUG(2, "dropped %d bytes looking for new frame", drop);
 	}
 
 	if(frame->scanstate == ScanState_Scanning)
 		return;
 		
-	/* Try to move data from queue into frame buffer 
+	/* Try to move data from queue into frame buffer
 	 * We get data in blocks of 384 bytes made up of:
 	 * 256 Y, 64 U, 64 V.
 	 * This needs to be written out as a Y plane, a U plane and a V plane.
 	 */
 		
-	while ( frame->curline < maxline && (n = RingQueue_GetLength(&uvd->dp)) >= 384) {
+	while ( frame->curline < maxline && (RingQueue_GetLength(&uvd->dp) >= 384)) {
 		/* Y */
 		RingQueue_Dequeue(&uvd->dp, frame->data + (frame->curline * 256), 256);
 		/* U */
@@ -497,8 +544,7 @@ static void konicawc_process_isoc(uvd_t *uvd, usbvideo_frame_t *frame)
 	}
 	/* See if we filled the frame */
 	if (frame->curline == maxline) {
-		if(debug > 5)
-			dbg("got whole frame");
+		DEBUG(5, "got whole frame");
 
 		frame->frameState = FrameState_Done_Hold;
 		frame->curline = 0;
@@ -510,10 +556,8 @@ static void konicawc_process_isoc(uvd_t *uvd, usbvideo_frame_t *frame)
 
 static int konicawc_calculate_fps(uvd_t *uvd)
 {
-	struct konicawc *t = uvd->user_data;
-	dbg("fps = %d", speed_to_fps[t->size][t->speed]/3);
-
-	return speed_to_fps[t->size][t->speed]/3;
+	struct konicawc *cam = uvd->user_data;
+	return spd_to_fps[cam->size][cam->speed]/3;
 }
 
 
@@ -550,10 +594,10 @@ static void konicawc_configure_video(uvd_t *uvd)
 	uvd->vcap.type = VID_TYPE_CAPTURE;
 	uvd->vcap.channels = 1;
 	uvd->vcap.audios = 0;
-	uvd->vcap.minwidth = camera_sizes[cam->size][0];
-	uvd->vcap.minheight = camera_sizes[cam->size][1];
-	uvd->vcap.maxwidth = camera_sizes[cam->size][0];
-	uvd->vcap.maxheight = camera_sizes[cam->size][1];
+	uvd->vcap.minwidth = camera_sizes[cam->size].width;
+	uvd->vcap.minheight = camera_sizes[cam->size].height;
+	uvd->vcap.maxwidth = camera_sizes[cam->size].width;
+	uvd->vcap.maxheight = camera_sizes[cam->size].height;
 
 	memset(&uvd->vchan, 0, sizeof(uvd->vchan));
 	uvd->vchan.flags = 0 ;
@@ -563,15 +607,14 @@ static void konicawc_configure_video(uvd_t *uvd)
 	strcpy(uvd->vchan.name, "Camera");
 
 	/* Talk to device */
-	dbg("device init");
+	DEBUG(1, "device init");
 	if(!konicawc_get_misc(uvd, 0x3, 0, 0x10, buf, 2))
-		dbg("3,10 -> %2.2x %2.2x", buf[0], buf[1]);
+		DEBUG(2, "3,10 -> %2.2x %2.2x", buf[0], buf[1]);
 	if(!konicawc_get_misc(uvd, 0x3, 0, 0x10, buf, 2))
-		dbg("3,10 -> %2.2x %2.2x", buf[0], buf[1]);
+		DEBUG(2, "3,10 -> %2.2x %2.2x", buf[0], buf[1]);
 	if(konicawc_set_misc(uvd, 0x2, 0, 0xd))
-		dbg("2,0,d failed");
-	dbg("setting initial values");
-
+		DEBUG(2, "2,0,d failed");
+	DEBUG(1, "setting initial values");
 }
 
 
@@ -582,8 +625,7 @@ static void *konicawc_probe(struct usb_device *dev, unsigned int ifnum, const st
 	int actInterface=-1, inactInterface=-1, maxPS=0;
 	unsigned char video_ep = 0;
 	
-	if (debug >= 1)
-		dbg("konicawc_probe(%p,%u.)", dev, ifnum);
+	DEBUG(1, "konicawc_probe(%p,%u.)", dev, ifnum);
 
 	/* We don't handle multi-config cameras */
 	if (dev->descriptor.bNumConfigurations != 1)
@@ -594,10 +636,8 @@ static void *konicawc_probe(struct usb_device *dev, unsigned int ifnum, const st
 
 	/* Validate found interface: must have one ISO endpoint */
 	nas = dev->actconfig->interface[ifnum].num_altsetting;
-	if (debug > 0)
-		info("Number of alternate settings=%d.", nas);
-	if (nas < 8) {
-		err("Too few alternate settings for this camera!");
+	if (nas != 8) {
+		err("Incorrect number of alternate settings (%d) for this camera!", nas);
 		return NULL;
 	}
 	/* Validate all alternate settings */
@@ -612,7 +652,7 @@ static void *konicawc_probe(struct usb_device *dev, unsigned int ifnum, const st
 			return NULL;
 		}
 		endpoint = &interface->endpoint[1];
-		dbg("found endpoint: addr: 0x%2.2x maxps = 0x%4.4x",
+		DEBUG(1, "found endpoint: addr: 0x%2.2x maxps = 0x%4.4x",
 		    endpoint->bEndpointAddress, endpoint->wMaxPacketSize);
 		if (video_ep == 0)
 			video_ep = endpoint->bEndpointAddress;
@@ -636,22 +676,20 @@ static void *konicawc_probe(struct usb_device *dev, unsigned int ifnum, const st
 				return NULL;
 			}
 		} else {
-			if (i == speed_to_interface[speed]) {
+			if (i == spd_to_iface[speed]) {
 				/* This one is the requested one */
 				actInterface = i;
-				maxPS = endpoint->wMaxPacketSize;
-				if (debug > 0) {
-					info("Selecting requested active setting=%d. maxPS=%d.",
-					     i, maxPS);
-				}
 			}
 		}
+		if(endpoint->wMaxPacketSize > maxPS)
+			maxPS = endpoint->wMaxPacketSize;
 	}
 	if(actInterface == -1) {
 		err("Cant find required endpoint");
 		return NULL;
 	}
 
+	DEBUG(1, "Selecting requested active setting=%d. maxPS=%d.", actInterface, maxPS);
 
 	/* Code below may sleep, need to lock module while we are here */
 	MOD_INC_USE_COUNT;
@@ -670,26 +708,10 @@ static void *konicawc_probe(struct usb_device *dev, unsigned int ifnum, const st
 			}
 		}
 		cam->speed = speed;
-		switch(size) {
-		case SIZE_160X136:
-		default:
-			cam->height = 136;
-			cam->width = 160;
-			cam->size = SIZE_160X136;
-			break;
-
-		case SIZE_176X144:
-			cam->height = 144;
-			cam->width = 176;
-			cam->size = SIZE_176X144;
-			break;
-
-		case SIZE_320X240:
-			cam->height = 240;
-			cam->width = 320;
-			cam->size = SIZE_320X240;
-			break;
-		}
+		RESTRICT_TO_RANGE(size, SIZE_160X120, SIZE_320X240);
+		cam->width = camera_sizes[size].width;
+		cam->height = camera_sizes[size].height;
+		cam->size = size;
 
 		uvd->flags = 0;
 		uvd->debug = debug;
@@ -773,9 +795,9 @@ MODULE_LICENSE("GPL");
 MODULE_AUTHOR("Simon Evans <spse@secret.org.uk>");
 MODULE_DESCRIPTION(DRIVER_DESC);
 MODULE_PARM(speed, "i");
-MODULE_PARM_DESC(speed, "FPS speed: 0 (slowest) - 6 (fastest)");
+MODULE_PARM_DESC(speed, "Initial speed: 0 (slowest) - 6 (fastest)");
 MODULE_PARM(size, "i");
-MODULE_PARM_DESC(size, "Frame Size 0: 160x136 1: 176x144 2: 320x240");
+MODULE_PARM_DESC(size, "Initial Size 0: 160x120 1: 160x136 2: 176x144 3: 320x240");
 MODULE_PARM(brightness, "i");
 MODULE_PARM_DESC(brightness, "Initial brightness 0 - 108");
 MODULE_PARM(contrast, "i");
@@ -786,7 +808,11 @@ MODULE_PARM(sharpness, "i");
 MODULE_PARM_DESC(sharpness, "Initial brightness 0 - 108");
 MODULE_PARM(whitebal, "i");
 MODULE_PARM_DESC(whitebal, "Initial white balance 0 - 363");
+
+#ifdef CONFIG_USB_DEBUG
 MODULE_PARM(debug, "i");
 MODULE_PARM_DESC(debug, "Debug level: 0-9 (default=0)");
+#endif
+
 module_init(konicawc_init);
 module_exit(konicawc_cleanup);

drivers/usb/misc/tiglusb.c

@@ -41,10 +41,6 @@
 #include <linux/ticable.h>
 #include "tiglusb.h"
 
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
-# define minor(x) MINOR(x)
-#endif
-
 /*
  * Version Information
  */

drivers/usb/net/cdc-ether.c

-// Portions of this file taken from 
+// Portions of this file taken from
 // Petko Manolov - Petkan (petkan@dce.bg)
 // from his driver pegasus.c
 
@@ -1170,23 +1170,20 @@ static void * CDCEther_probe( struct usb_device *usb, unsigned int ifnum,
 	if (rc)	{
 		// Nope we couldn't find one we liked.
 		// This device was not meant for us to control.
-		kfree( ether_dev );
-		return	NULL;
+		goto error_all;
 	}
 
-	// Now that we FOUND a configuration. let's try to make the 
+	// Now that we FOUND a configuration. let's try to make the
 	// device go into it.
 	if ( usb_set_configuration( usb, ether_dev->bConfigurationValue ) ) {
 		err("usb_set_configuration() failed");
-		kfree( ether_dev );
-		return NULL;
+		goto error_all;
 	}
 
 	// Now set the communication interface up as required.
 	if (usb_set_interface(usb, ether_dev->comm_bInterfaceNumber, ether_dev->comm_bAlternateSetting)) {
 		err("usb_set_interface() failed");
-		kfree( ether_dev );
-		return NULL;
+		goto error_all;
 	}
 
 	// Only turn traffic on right now if we must...
@@ -1194,23 +1191,21 @@ static void * CDCEther_probe( struct usb_device *usb, unsigned int ifnum,
 		// We found an alternate setting for the data
 		// interface that allows us to turn off traffic.
 		// We should use it.
-		if (usb_set_interface( usb, 
-		                       ether_dev->data_bInterfaceNumber, 
+		if (usb_set_interface( usb,
+		                       ether_dev->data_bInterfaceNumber,
 		                       ether_dev->data_bAlternateSetting_without_traffic)) {
 			err("usb_set_interface() failed");
-			kfree( ether_dev );
-			return NULL;
+			goto error_all;
 		}
 	} else	{
 		// We didn't find an alternate setting for the data
 		// interface that would let us turn off traffic.
 		// Oh well, let's go ahead and do what we must...
-		if (usb_set_interface( usb, 
-		                       ether_dev->data_bInterfaceNumber, 
+		if (usb_set_interface( usb,
+		                       ether_dev->data_bInterfaceNumber,
 		                       ether_dev->data_bAlternateSetting_with_traffic)) {
 			err("usb_set_interface() failed");
-			kfree( ether_dev );
-			return NULL;
+			goto error_all;
 		}
 	}
 
@@ -1220,8 +1215,7 @@ static void * CDCEther_probe( struct usb_device *usb, unsigned int ifnum,
 		// Hmm...  The kernel is not sharing today...
 		// Fine, we didn't want it anyway...
 		err( "Unable to initialize ethernet device" );
-		kfree( ether_dev );
-		return	NULL;
+		goto error_all;
 	}
 
 	// Now that we have an ethernet device, let's set it up
@@ -1241,7 +1235,7 @@ static void * CDCEther_probe( struct usb_device *usb, unsigned int ifnum,
 	// We'll keep track of this information for later...
 	ether_dev->usb = usb;
 	ether_dev->net = net;
-	
+
 	// and don't forget the MAC address.
 	set_ethernet_addr( ether_dev );
 
@@ -1249,12 +1243,12 @@ static void * CDCEther_probe( struct usb_device *usb, unsigned int ifnum,
 	log_device_info( ether_dev );
 
 	// I claim this interface to be a CDC Ethernet Networking device
-	usb_driver_claim_interface( &CDCEther_driver, 
-	                            &(usb->config[ether_dev->configuration_num].interface[ether_dev->comm_interface]), 
+	usb_driver_claim_interface( &CDCEther_driver,
+	                            &(usb->config[ether_dev->configuration_num].interface[ether_dev->comm_interface]),
 	                            ether_dev );
 	// I claim this interface to be a CDC Ethernet Networking device
-	usb_driver_claim_interface( &CDCEther_driver, 
-	                            &(usb->config[ether_dev->configuration_num].interface[ether_dev->data_interface]), 
+	usb_driver_claim_interface( &CDCEther_driver,
+	                            &(usb->config[ether_dev->configuration_num].interface[ether_dev->data_interface]),
 	                            ether_dev );
 
 	// Does this REALLY do anything???
@@ -1265,6 +1259,14 @@ static void * CDCEther_probe( struct usb_device *usb, unsigned int ifnum,
 
 	// Okay, we are finally done...
 	return NULL;
+
+	// bailing out with our tail between our knees
+error_all:
+	usb_free_urb(ether_dev->tx_urb);
+	usb_free_urb(ether_dev->rx_urb);
+	usb_free_urb(ether_dev->intr_urb);
+	kfree( ether_dev );
+	return	NULL;
 }
 
 

drivers/usb/storage/scsiglue.c

@@ -147,7 +147,8 @@ static int queuecommand( Scsi_Cmnd *srb , void (*done)(Scsi_Cmnd *))
 	srb->host_scribble = (unsigned char *)us;
 
 	/* enqueue the command */
-	BUG_ON(atomic_read(&us->sm_state) != US_STATE_IDLE || us->srb != NULL);
+	BUG_ON(atomic_read(&us->sm_state) != US_STATE_IDLE);
+	BUG_ON(us->srb != NULL);
 	srb->scsi_done = done;
 	us->srb = srb;
 

drivers/usb/storage/usb.h

@@ -203,16 +203,9 @@ extern void fill_inquiry_response(struct us_data *us,
 
 /* The scsi_lock() and scsi_unlock() macros protect the sm_state and the
  * single queue element srb for write access */
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,3)
 #define scsi_unlock(host)	spin_unlock_irq(host->host_lock)
 #define scsi_lock(host)		spin_lock_irq(host->host_lock)
 
 #define sg_address(psg)		(page_address((psg)->page) + (psg)->offset)
-#else
-#define scsi_unlock(host)	spin_unlock_irq(&io_request_lock)
-#define scsi_lock(host)		spin_lock_irq(&io_request_lock)
-
-#define sg_address(psg)		((psg)->address)
-#endif
 
 #endif

fs/block_dev.c

@@ -874,7 +874,7 @@ struct file_operations def_blk_fops = {
 	release:	blkdev_close,
 	llseek:		block_llseek,
 	read:		generic_file_read,
-	write:		generic_file_write,
+	write:		generic_file_write_nolock,
 	mmap:		generic_file_mmap,
 	fsync:		block_fsync,
 	ioctl:		blkdev_ioctl,

fs/fat/file.c

@@ -21,6 +21,7 @@ struct file_operations fat_file_operations = {
 	write:		fat_file_write,
 	mmap:		generic_file_mmap,
 	fsync:		file_fsync,
+	sendfile:	generic_file_sendfile,
 };
 
 struct inode_operations fat_file_inode_operations = {

fs/partitions/check.c

@@ -137,11 +137,6 @@ char *disk_name (struct gendisk *hd, int minor, char *buf)
 			sprintf(s, "%s%d", "md", unit);
 			maj = s;
 			break;
-		case COMPAQ_CISS_MAJOR ... COMPAQ_CISS_MAJOR+7:
-			sprintf(s, "cciss/c%dd%d",
-				hd->major - COMPAQ_CISS_MAJOR, unit);
-			maj = s;
-			break;
 		case ATARAID_MAJOR:
 			sprintf(s, "ataraid/d%d", unit);
 			maj = s;

include/asm-alpha/rtc.h

+#ifndef _ALPHA_RTC_H
+#define _ALPHA_RTC_H
+
+/*
+ * Alpha uses the default access methods for the RTC.
+ */
+
+#include <asm-generic/rtc.h>
+
+#endif

include/asm-generic/rtc.h

+/* 
+ * inclue/asm-generic/rtc.h
+ *
+ * Author: Tom Rini <trini@mvista.com>
+ *
+ * Based on:
+ * drivers/char/rtc.c
+ *
+ * Please read the COPYING file for all license details.
+ */
+
+#ifndef __ASM_RTC_H__
+#define __ASM_RTC_H__
+
+#ifdef __KERNEL__
+
+#include <linux/mc146818rtc.h>
+#include <linux/rtc.h>
+
+#define RTC_PIE 0x40		/* periodic interrupt enable */
+#define RTC_AIE 0x20		/* alarm interrupt enable */
+#define RTC_UIE 0x10		/* update-finished interrupt enable */
+
+extern void gen_rtc_interrupt(unsigned long);
+
+/* some dummy definitions */
+#define RTC_SQWE 0x08		/* enable square-wave output */
+#define RTC_DM_BINARY 0x04	/* all time/date values are BCD if clear */
+#define RTC_24H 0x02		/* 24 hour mode - else hours bit 7 means pm */
+#define RTC_DST_EN 0x01	        /* auto switch DST - works f. USA only */
+
+/*
+ * Returns true if a clock update is in progress
+ */
+static inline unsigned char rtc_is_updating(void)
+{
+	unsigned char uip;
+
+	spin_lock_irq(&rtc_lock);
+	uip = (CMOS_READ(RTC_FREQ_SELECT) & RTC_UIP);
+	spin_unlock_irq(&rtc_lock);
+	return uip;
+}
+
+static inline void get_rtc_time(struct rtc_time *time)
+{
+	unsigned long uip_watchdog = jiffies;
+	unsigned char ctrl;
+#ifdef CONFIG_DECSTATION
+	unsigned int real_year;
+#endif
+
+	/*
+	 * read RTC once any update in progress is done. The update
+	 * can take just over 2ms. We wait 10 to 20ms. There is no need to
+	 * to poll-wait (up to 1s - eeccch) for the falling edge of RTC_UIP.
+	 * If you need to know *exactly* when a second has started, enable
+	 * periodic update complete interrupts, (via ioctl) and then 
+	 * immediately read /dev/rtc which will block until you get the IRQ.
+	 * Once the read clears, read the RTC time (again via ioctl). Easy.
+	 */
+
+	if (rtc_is_updating() != 0)
+		while (jiffies - uip_watchdog < 2*HZ/100) {
+			barrier();
+			cpu_relax();
+		}
+
+	/*
+	 * Only the values that we read from the RTC are set. We leave
+	 * tm_wday, tm_yday and tm_isdst untouched. Even though the
+	 * RTC has RTC_DAY_OF_WEEK, we ignore it, as it is only updated
+	 * by the RTC when initially set to a non-zero value.
+	 */
+	spin_lock_irq(&rtc_lock);
+	time->tm_sec = CMOS_READ(RTC_SECONDS);
+	time->tm_min = CMOS_READ(RTC_MINUTES);
+	time->tm_hour = CMOS_READ(RTC_HOURS);
+	time->tm_mday = CMOS_READ(RTC_DAY_OF_MONTH);
+	time->tm_mon = CMOS_READ(RTC_MONTH);
+	time->tm_year = CMOS_READ(RTC_YEAR);
+#ifdef CONFIG_DECSTATION
+	real_year = CMOS_READ(RTC_DEC_YEAR);
+#endif
+	ctrl = CMOS_READ(RTC_CONTROL);
+	spin_unlock_irq(&rtc_lock);
+
+	if (!(ctrl & RTC_DM_BINARY) || RTC_ALWAYS_BCD)
+	{
+		BCD_TO_BIN(time->tm_sec);
+		BCD_TO_BIN(time->tm_min);
+		BCD_TO_BIN(time->tm_hour);
+		BCD_TO_BIN(time->tm_mday);
+		BCD_TO_BIN(time->tm_mon);
+		BCD_TO_BIN(time->tm_year);
+	}
+
+#ifdef CONFIG_DECSTATION
+	time->tm_year += real_year - 72;
+#endif
+
+	/*
+	 * Account for differences between how the RTC uses the values
+	 * and how they are defined in a struct rtc_time;
+	 */
+	if (time->tm_year <= 69)
+		time->tm_year += 100;
+
+	time->tm_mon--;
+}
+
+/* Set the current date and time in the real time clock. */
+static inline int set_rtc_time(struct rtc_time *time)
+{
+	unsigned char mon, day, hrs, min, sec;
+	unsigned char save_control, save_freq_select;
+	unsigned int yrs;
+#ifdef CONFIG_DECSTATION
+	unsigned int real_yrs, leap_yr;
+#endif
+
+	yrs = time->tm_year;
+	mon = time->tm_mon + 1;   /* tm_mon starts at zero */
+	day = time->tm_mday;
+	hrs = time->tm_hour;
+	min = time->tm_min;
+	sec = time->tm_sec;
+
+	if (yrs > 255)	/* They are unsigned */
+		return -EINVAL;
+
+	spin_lock_irq(&rtc_lock);
+#ifdef CONFIG_DECSTATION
+	real_yrs = yrs;
+	leap_yr = ((!((yrs + 1900) % 4) && ((yrs + 1900) % 100)) ||
+			!((yrs + 1900) % 400));
+	yrs = 72;
+
+	/*
+	 * We want to keep the year set to 73 until March
+	 * for non-leap years, so that Feb, 29th is handled
+	 * correctly.
+	 */
+	if (!leap_yr && mon < 3) {
+		real_yrs--;
+		yrs = 73;
+	}
+#endif
+	/* These limits and adjustments are independant of
+	 * whether the chip is in binary mode or not.
+	 */
+	if (yrs > 169) {
+		spin_unlock_irq(&rtc_lock);
+		return -EINVAL;
+	}
+
+	if (yrs >= 100)
+		yrs -= 100;
+
+	if (!(CMOS_READ(RTC_CONTROL) & RTC_DM_BINARY)
+	    || RTC_ALWAYS_BCD) {
+		BIN_TO_BCD(sec);
+		BIN_TO_BCD(min);
+		BIN_TO_BCD(hrs);
+		BIN_TO_BCD(day);
+		BIN_TO_BCD(mon);
+		BIN_TO_BCD(yrs);
+	}
+
+	save_control = CMOS_READ(RTC_CONTROL);
+	CMOS_WRITE((save_control|RTC_SET), RTC_CONTROL);
+	save_freq_select = CMOS_READ(RTC_FREQ_SELECT);
+	CMOS_WRITE((save_freq_select|RTC_DIV_RESET2), RTC_FREQ_SELECT);
+
+#ifdef CONFIG_DECSTATION
+	CMOS_WRITE(real_yrs, RTC_DEC_YEAR);
+#endif
+	CMOS_WRITE(yrs, RTC_YEAR);
+	CMOS_WRITE(mon, RTC_MONTH);
+	CMOS_WRITE(day, RTC_DAY_OF_MONTH);
+	CMOS_WRITE(hrs, RTC_HOURS);
+	CMOS_WRITE(min, RTC_MINUTES);
+	CMOS_WRITE(sec, RTC_SECONDS);
+
+	CMOS_WRITE(save_control, RTC_CONTROL);
+	CMOS_WRITE(save_freq_select, RTC_FREQ_SELECT);
+
+	spin_unlock_irq(&rtc_lock);
+
+	return 0;
+}
+
+static inline unsigned int get_rtc_ss(void)
+{
+	struct rtc_time h;
+
+	get_rtc_time(&h);
+	return h.tm_sec;
+}
+
+static inline int get_rtc_pll(struct rtc_pll_info *pll)
+{
+	return -EINVAL;
+}
+static inline int set_rtc_pll(struct rtc_pll_info *pll)
+{
+	return -EINVAL;
+}
+
+#endif /* __KERNEL__ */
+#endif /* __ASM_RTC_H__ */

include/asm-i386/rtc.h

+#ifndef _I386_RTC_H
+#define _I386_RTC_H
+
+/*
+ * x86 uses the default access methods for the RTC.
+ */
+
+#include <asm-generic/rtc.h>
+
+#endif

include/asm-parisc/rtc.h

+/* 
+ * inclue/asm-parisc/rtc.h
+ *
+ * Copyright 2002 Randolph CHung <tausq@debian.org>
+ *
+ * Based on: include/asm-ppc/rtc.h and the genrtc driver in the
+ * 2.4 parisc linux tree
+ */
+
+#ifndef __ASM_RTC_H__
+#define __ASM_RTC_H__
+
+#ifdef __KERNEL__
+
+#include <linux/rtc.h>
+
+#include <asm/pdc.h>
+
+#define SECS_PER_HOUR   (60 * 60)
+#define SECS_PER_DAY    (SECS_PER_HOUR * 24)
+
+
+#define RTC_PIE 0x40		/* periodic interrupt enable */
+#define RTC_AIE 0x20		/* alarm interrupt enable */
+#define RTC_UIE 0x10		/* update-finished interrupt enable */
+
+extern void gen_rtc_interrupt(unsigned long);
+
+/* some dummy definitions */
+#define RTC_SQWE 0x08		/* enable square-wave output */
+#define RTC_DM_BINARY 0x04	/* all time/date values are BCD if clear */
+#define RTC_24H 0x02		/* 24 hour mode - else hours bit 7 means pm */
+#define RTC_DST_EN 0x01	        /* auto switch DST - works f. USA only */
+
+# define __isleap(year) \
+  ((year) % 4 == 0 && ((year) % 100 != 0 || (year) % 400 == 0))
+
+/* How many days come before each month (0-12).  */
+static const unsigned short int __mon_yday[2][13] =
+{
+	/* Normal years.  */
+	{ 0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334, 365 },
+	/* Leap years.  */
+	{ 0, 31, 60, 91, 121, 152, 182, 213, 244, 274, 305, 335, 366 }
+};
+
+static int get_rtc_time(struct rtc_time *wtime)
+{
+	struct pdc_tod tod_data;
+	long int days, rem, y;
+	const unsigned short int *ip;
+
+	if(pdc_tod_read(&tod_data) < 0)
+		return -1;
+
+	
+	// most of the remainder of this function is:
+//	Copyright (C) 1991, 1993, 1997, 1998 Free Software Foundation, Inc.
+//	This was originally a part of the GNU C Library.
+//      It is distributed under the GPL, and was swiped from offtime.c
+
+
+	days = tod_data.tod_sec / SECS_PER_DAY;
+	rem = tod_data.tod_sec % SECS_PER_DAY;
+
+	wtime->tm_hour = rem / SECS_PER_HOUR;
+	rem %= SECS_PER_HOUR;
+	wtime->tm_min = rem / 60;
+	wtime->tm_sec = rem % 60;
+
+	y = 1970;
+	
+#define DIV(a, b) ((a) / (b) - ((a) % (b) < 0))
+#define LEAPS_THRU_END_OF(y) (DIV (y, 4) - DIV (y, 100) + DIV (y, 400))
+
+	while (days < 0 || days >= (__isleap (y) ? 366 : 365))
+	{
+		/* Guess a corrected year, assuming 365 days per year.  */
+		long int yg = y + days / 365 - (days % 365 < 0);
+
+		/* Adjust DAYS and Y to match the guessed year.  */
+		days -= ((yg - y) * 365
+			 + LEAPS_THRU_END_OF (yg - 1)
+			 - LEAPS_THRU_END_OF (y - 1));
+		y = yg;
+	}
+	wtime->tm_year = y - 1900;
+
+	ip = __mon_yday[__isleap(y)];
+	for (y = 11; days < (long int) ip[y]; --y)
+		continue;
+	days -= ip[y];
+	wtime->tm_mon = y;
+	wtime->tm_mday = days + 1;
+	
+	return 0;
+}
+
+static int set_rtc_time(struct rtc_time *wtime)
+{
+	u_int32_t secs;
+
+	secs = mktime(wtime->tm_year + 1900, wtime->tm_mon + 1, wtime->tm_mday, 
+		      wtime->tm_hour, wtime->tm_min, wtime->tm_sec);
+
+	if(pdc_tod_set(secs, 0) < 0)
+		return -1;
+	else
+		return 0;
+
+}
+
+static inline unsigned int get_rtc_ss(void)
+{
+	struct rtc_time h;
+
+	get_rtc_time(&h);
+	return h.tm_sec;
+}
+
+static inline int get_rtc_pll(struct rtc_pll_info *pll)
+{
+	return -EINVAL;
+}
+static inline int set_rtc_pll(struct rtc_pll_info *pll)
+{
+	return -EINVAL;
+}
+
+#endif /* __KERNEL__ */
+#endif /* __ASM_RTC_H__ */

include/asm-ppc/rtc.h

+/* 
+ * inclue/asm-ppc/rtc.h
+ *
+ * Copyright 2002 MontaVista Software Inc.
+ * Author: Tom Rini <trini@mvista.com>
+ *
+ * Based on:
+ * include/asm-m68k/rtc.h
+ *
+ * Copyright Richard Zidlicky
+ * implementation details for genrtc/q40rtc driver
+ *
+ * And the old drivers/macintosh/rtc.c which was heavily based on:
+ * Linux/SPARC Real Time Clock Driver
+ * Copyright (C) 1996 Thomas K. Dyas (tdyas@eden.rutgers.edu)
+ *
+ * With additional work by Paul Mackerras and Franz Sirl.
+ */
+/* permission is hereby granted to copy, modify and redistribute this code
+ * in terms of the GNU Library General Public License, Version 2 or later,
+ * at your option.
+ */
+
+#ifndef __ASM_RTC_H__
+#define __ASM_RTC_H__
+
+#ifdef __KERNEL__
+
+#include <linux/rtc.h>
+
+#include <asm/machdep.h>
+#include <asm/time.h>
+
+#define RTC_PIE 0x40		/* periodic interrupt enable */
+#define RTC_AIE 0x20		/* alarm interrupt enable */
+#define RTC_UIE 0x10		/* update-finished interrupt enable */
+
+extern void gen_rtc_interrupt(unsigned long);
+
+/* some dummy definitions */
+#define RTC_SQWE 0x08		/* enable square-wave output */
+#define RTC_DM_BINARY 0x04	/* all time/date values are BCD if clear */
+#define RTC_24H 0x02		/* 24 hour mode - else hours bit 7 means pm */
+#define RTC_DST_EN 0x01	        /* auto switch DST - works f. USA only */
+
+static inline void get_rtc_time(struct rtc_time *time)
+{
+	if (ppc_md.get_rtc_time) {
+		unsigned long nowtime;
+
+		nowtime = (ppc_md.get_rtc_time)();
+
+		to_tm(nowtime, time);
+
+		time->tm_year -= 1900;
+		time->tm_mon -= 1; /* Make sure userland has a 0-based month */
+	}
+}
+
+/* Set the current date and time in the real time clock. */
+static inline void set_rtc_time(struct rtc_time *time)
+{
+	if (ppc_md.get_rtc_time) {
+		unsigned long nowtime;
+
+		nowtime = mktime(time->tm_year+1900, time->tm_mon+1,
+				time->tm_mday, time->tm_hour, time->tm_min,
+				time->tm_sec);
+
+		(ppc_md.set_rtc_time)(nowtime);
+
+		return 0;
+	} else
+		return -EINVAL;
+}
+
+static inline unsigned int get_rtc_ss(void)
+{
+	struct rtc_time h;
+
+	get_rtc_time(&h);
+	return h.tm_sec;
+}
+
+static inline int get_rtc_pll(struct rtc_pll_info *pll)
+{
+	return -EINVAL;
+}
+static inline int set_rtc_pll(struct rtc_pll_info *pll)
+{
+	return -EINVAL;
+}
+
+#endif /* __KERNEL__ */
+#endif /* __ASM_RTC_H__ */

include/linux/apm_bios.h

@@ -21,8 +21,7 @@ typedef unsigned short	apm_eventinfo_t;
 
 #ifdef __KERNEL__
 
-#define APM_40		(GDT_ENTRY_APMBIOS_BASE * 8)
-#define APM_CS		(APM_BASE + 8)
+#define APM_CS		(GDT_ENTRY_APMBIOS_BASE * 8)
 #define APM_CS_16	(APM_CS + 8)
 #define APM_DS		(APM_CS_16 + 8)
 

include/linux/fs.h

@@ -1241,6 +1241,7 @@ extern int generic_file_mmap(struct file *, struct vm_area_struct *);
 extern int file_read_actor(read_descriptor_t * desc, struct page *page, unsigned long offset, unsigned long size);
 extern ssize_t generic_file_read(struct file *, char *, size_t, loff_t *);
 extern ssize_t generic_file_write(struct file *, const char *, size_t, loff_t *);
+extern ssize_t generic_file_write_nolock(struct file *, const char *, size_t, loff_t *);
 extern ssize_t generic_file_sendfile(struct file *, struct file *, loff_t *, size_t);
 extern void do_generic_file_read(struct file *, loff_t *, read_descriptor_t *, read_actor_t);
 ssize_t generic_file_direct_IO(int rw, struct inode *inode, char *buf,

include/linux/init_task.h

@@ -5,34 +5,34 @@
 
 #define INIT_FILES \
 { 							\
-	count:		ATOMIC_INIT(1), 		\
-	file_lock:	RW_LOCK_UNLOCKED, 		\
-	max_fds:	NR_OPEN_DEFAULT, 		\
-	max_fdset:	__FD_SETSIZE, 			\
-	next_fd:	0, 				\
-	fd:		&init_files.fd_array[0], 	\
-	close_on_exec:	&init_files.close_on_exec_init, \
-	open_fds:	&init_files.open_fds_init, 	\
-	close_on_exec_init: { { 0, } }, 		\
-	open_fds_init:	{ { 0, } }, 			\
-	fd_array:	{ NULL, } 			\
+	.count		= ATOMIC_INIT(1), 		\
+	.file_lock	= RW_LOCK_UNLOCKED, 		\
+	.max_fds	= NR_OPEN_DEFAULT, 		\
+	.max_fdset	= __FD_SETSIZE, 		\
+	.next_fd	= 0, 				\
+	.fd		= &init_files.fd_array[0], 	\
+	.close_on_exec	= &init_files.close_on_exec_init, \
+	.open_fds	= &init_files.open_fds_init, 	\
+	.close_on_exec_init = { { 0, } }, 		\
+	.open_fds_init	= { { 0, } }, 			\
+	.fd_array	= { NULL, } 			\
 }
 
 #define INIT_MM(name) \
 {			 				\
-	mm_rb:		RB_ROOT,			\
-	pgd:		swapper_pg_dir, 		\
-	mm_users:	ATOMIC_INIT(2), 		\
-	mm_count:	ATOMIC_INIT(1), 		\
-	mmap_sem:	__RWSEM_INITIALIZER(name.mmap_sem), \
-	page_table_lock: SPIN_LOCK_UNLOCKED, 		\
-	mmlist:		LIST_HEAD_INIT(name.mmlist),	\
+	.mm_rb		= RB_ROOT,			\
+	.pgd		= swapper_pg_dir, 		\
+	.mm_users	= ATOMIC_INIT(2), 		\
+	.mm_count	= ATOMIC_INIT(1), 		\
+	.mmap_sem	= __RWSEM_INITIALIZER(name.mmap_sem), \
+	.page_table_lock =  SPIN_LOCK_UNLOCKED, 	\
+	.mmlist		= LIST_HEAD_INIT(name.mmlist),	\
 }
 
 #define INIT_SIGNALS {	\
-	count:		ATOMIC_INIT(1), 		\
-	action:		{ {{0,}}, }, 			\
-	siglock:	SPIN_LOCK_UNLOCKED 		\
+	.count		= ATOMIC_INIT(1), 		\
+	.action		= { {{0,}}, }, 			\
+	.siglock	= SPIN_LOCK_UNLOCKED 		\
 }
 
 /*
@@ -41,45 +41,45 @@
  */
 #define INIT_TASK(tsk)	\
 {									\
-    state:		0,						\
-    thread_info:	&init_thread_info,				\
-    flags:		0,						\
-    lock_depth:		-1,						\
-    prio:		MAX_PRIO-20,					\
-    static_prio:	MAX_PRIO-20,					\
-    policy:		SCHED_NORMAL,					\
-    cpus_allowed:	-1,						\
-    mm:			NULL,						\
-    active_mm:		&init_mm,					\
-    run_list:		LIST_HEAD_INIT(tsk.run_list),			\
-    time_slice:		HZ,						\
-    tasks:		LIST_HEAD_INIT(tsk.tasks),			\
-    real_parent:	&tsk,						\
-    parent:		&tsk,						\
-    children:		LIST_HEAD_INIT(tsk.children),			\
-    sibling:		LIST_HEAD_INIT(tsk.sibling),			\
-    thread_group:	LIST_HEAD_INIT(tsk.thread_group),		\
-    wait_chldexit:	__WAIT_QUEUE_HEAD_INITIALIZER(tsk.wait_chldexit),\
-    real_timer:		{						\
-	function:		it_real_fn				\
-    },									\
-    cap_effective:	CAP_INIT_EFF_SET,				\
-    cap_inheritable:	CAP_INIT_INH_SET,				\
-    cap_permitted:	CAP_FULL_SET,					\
-    keep_capabilities:	0,						\
-    rlim:		INIT_RLIMITS,					\
-    user:		INIT_USER,					\
-    comm:		"swapper",					\
-    thread:		INIT_THREAD,					\
-    fs:			&init_fs,					\
-    files:		&init_files,					\
-    sigmask_lock:	SPIN_LOCK_UNLOCKED,				\
-    sig:		&init_signals,					\
-    pending:		{ NULL, &tsk.pending.head, {{0}}},		\
-    blocked:		{{0}},						\
-    alloc_lock:		SPIN_LOCK_UNLOCKED,				\
-    switch_lock:	SPIN_LOCK_UNLOCKED,				\
-    journal_info:	NULL,						\
+	.state		= 0,						\
+	.thread_info	= &init_thread_info,				\
+	.flags		= 0,						\
+	.lock_depth	= -1,						\
+	.prio		= MAX_PRIO-20,					\
+	.static_prio	= MAX_PRIO-20,					\
+	.policy		= SCHED_NORMAL,					\
+	.cpus_allowed	= -1,						\
+	.mm		= NULL,						\
+	.active_mm	= &init_mm,					\
+	.run_list	= LIST_HEAD_INIT(tsk.run_list),			\
+	.time_slice	= HZ,						\
+	.tasks		= LIST_HEAD_INIT(tsk.tasks),			\
+	.real_parent	= &tsk,						\
+	.parent		= &tsk,						\
+	.children	= LIST_HEAD_INIT(tsk.children),			\
+	.sibling	= LIST_HEAD_INIT(tsk.sibling),			\
+	.thread_group	= LIST_HEAD_INIT(tsk.thread_group),		\
+	.wait_chldexit	= __WAIT_QUEUE_HEAD_INITIALIZER(tsk.wait_chldexit),\
+	.real_timer	= {						\
+		.function	= it_real_fn				\
+	},								\
+	.cap_effective	= CAP_INIT_EFF_SET,				\
+	.cap_inheritable = CAP_INIT_INH_SET,				\
+	.cap_permitted	= CAP_FULL_SET,					\
+	.keep_capabilities = 0,						\
+	.rlim		= INIT_RLIMITS,					\
+	.user		= INIT_USER,					\
+	.comm		= "swapper",					\
+	.thread		= INIT_THREAD,					\
+	.fs		= &init_fs,					\
+	.files		= &init_files,					\
+	.sigmask_lock	= SPIN_LOCK_UNLOCKED,				\
+	.sig		= &init_signals,				\
+	.pending	= { NULL, &tsk.pending.head, {{0}}},		\
+	.blocked	= {{0}},					\
+	.alloc_lock	= SPIN_LOCK_UNLOCKED,				\
+	.switch_lock	= SPIN_LOCK_UNLOCKED,				\
+	.journal_info	= NULL,						\
 }
 
 

include/linux/rtc.h

@@ -39,10 +39,32 @@ struct rtc_wkalrm {
 	struct rtc_time time;	/* time the alarm is set to */
 };
 
+/*
+ * Data structure to control PLL correction some better RTC feature
+ * pll_value is used to get or set current value of correction,
+ * the rest of the struct is used to query HW capabilities.
+ * This is modeled after the RTC used in Q40/Q60 computers but
+ * should be sufficiently flexible for other devices
+ *
+ * +ve pll_value means clock will run faster by
+ *   pll_value*pll_posmult/pll_clock
+ * -ve pll_value means clock will run slower by
+ *   pll_value*pll_negmult/pll_clock
+ */ 
+
+struct rtc_pll_info {
+	int pll_ctrl;       /* placeholder for fancier control */
+	int pll_value;      /* get/set correction value */
+	int pll_max;        /* max +ve (faster) adjustment value */
+	int pll_min;        /* max -ve (slower) adjustment value */
+	int pll_posmult;    /* factor for +ve corection */
+	int pll_negmult;    /* factor for -ve corection */
+	long pll_clock;     /* base PLL frequency */
+};
 
 /*
  * ioctl calls that are permitted to the /dev/rtc interface, if 
- * CONFIG_RTC/CONFIG_EFI_RTC was enabled.
+ * any of the RTC drivers are enabled.
  */
 
 #define RTC_AIE_ON	_IO('p', 0x01)	/* Alarm int. enable on		*/
@@ -65,6 +87,8 @@ struct rtc_wkalrm {
 
 #define RTC_WKALM_SET	_IOW('p', 0x0f, struct rtc_wkalrm)/* Set wakeup alarm*/
 #define RTC_WKALM_RD	_IOR('p', 0x10, struct rtc_wkalrm)/* Get wakeup alarm*/
+#define RTC_PLL_GET	_IOR('p', 0x11, struct rtc_pll_info)  /* Get PLL correction */
+#define RTC_PLL_SET	_IOW('p', 0x12, struct rtc_pll_info)  /* Set PLL correction */
 
 #ifdef __KERNEL__
 

include/linux/sunrpc/svcsock.h

@@ -30,6 +30,7 @@ struct svc_sock {
 #define	SK_TEMP		4			/* temp (TCP) socket */
 #define	SK_QUED		5			/* on serv->sk_sockets */
 #define	SK_DEAD		6			/* socket closed */
+#define	SK_CHNGBUF	7			/* need to change snd/rcv buffer sizes */
 
 	int			sk_reserved;	/* space on outq that is reserved */
 

include/linux/usb.h

@@ -734,6 +734,7 @@ extern void usb_deregister_dev(int num_minors, int start_minor);
  */
 #define URB_SHORT_NOT_OK	0x0001	/* report short reads as errors */
 #define USB_ISO_ASAP		0x0002	/* iso-only, urb->start_frame ignored */
+#define URB_NO_DMA_MAP		0x0004	/* urb->*_dma are valid on submit */
 #define USB_ASYNC_UNLINK	0x0008	/* usb_unlink_urb() returns asap */
 #define USB_NO_FSBR		0x0020	/* UHCI-specific */
 #define USB_ZERO_PACKET		0x0040	/* Finish bulk OUTs with short packet */
@@ -771,11 +772,15 @@ typedef void (*usb_complete_t)(struct urb *);
  * @transfer_flags: A variety of flags may be used to affect how URB
  *	submission, unlinking, or operation are handled.  Different
  *	kinds of URB can use different flags.
- * @transfer_buffer: For non-iso transfers, this identifies the buffer
- *	to (or from) which the I/O request will be performed.  This
- *	buffer must be suitable for DMA; allocate it with kmalloc()
+ * @transfer_buffer:  This identifies the buffer to (or from) which
+ * 	the I/O request will be performed (unless URB_NO_DMA_MAP is set).
+ *	This buffer must be suitable for DMA; allocate it with kmalloc()
  *	or equivalent.  For transfers to "in" endpoints, contents of
- *	this buffer will be modified.
+ *	this buffer will be modified.  This buffer is used for data
+ *	phases of control transfers.
+ * @transfer_dma: When transfer_flags includes URB_NO_DMA_MAP, the device
+ * 	driver is saying that it provided this DMA address, which the host
+ * 	controller driver should use instead of the transfer_buffer.
  * @transfer_buffer_length: How big is transfer_buffer.  The transfer may
  *	be broken up into chunks according to the current maximum packet
  *	size for the endpoint, which is a function of the configuration
@@ -789,6 +794,11 @@ typedef void (*usb_complete_t)(struct urb *);
  * @setup_packet: Only used for control transfers, this points to eight bytes
  *	of setup data.  Control transfers always start by sending this data
  *	to the device.  Then transfer_buffer is read or written, if needed.
+ *	(Not used when URB_NO_DMA_MAP is set.)
+ * @setup_dma: For control transfers with URB_NO_DMA_MAP set, the device
+ * 	driver has provided this DMA address for the setup packet.  The
+ * 	host controller driver should use instead of setup_buffer.
+ * 	If there is a data phase, its buffer is identified by transfer_dma.
  * @start_frame: Returns the initial frame for interrupt or isochronous
  *	transfers.
  * @number_of_packets: Lists the number of ISO transfer buffers.
@@ -811,6 +821,23 @@ typedef void (*usb_complete_t)(struct urb *);
  * are submitted using usb_submit_urb(), and pending requests may be canceled
  * using usb_unlink_urb().
  *
+ * Data Transfer Buffers:
+ *
+ * Normally drivers provide I/O buffers allocated with kmalloc() or otherwise
+ * taken from the general page pool.  That is provided by transfer_buffer
+ * (control requests also use setup_packet), and host controller drivers
+ * perform a dma mapping (and unmapping) for each buffer transferred.  Those
+ * mapping operations can be expensive on some platforms (such using a dma
+ * bounce buffer), although they're cheap on commodity x86 and ppc hardware.
+ *
+ * Alternatively, drivers may pass the URB_NO_DMA_MAP transfer flag, which
+ * tells the host controller driver that no such mapping is needed since
+ * the device driver is DMA-aware.  For example, they might allocate a DMA
+ * buffer with usb_buffer_alloc(), or call usb_buffer_map().
+ * When this transfer flag is provided, host controller drivers will use the
+ * dma addresses found in the transfer_dma and/or setup_dma fields rather than
+ * determing a dma address themselves.
+ *
  * Initialization:
  *
  * All URBs submitted must initialize dev, pipe,
@@ -818,10 +845,10 @@ typedef void (*usb_complete_t)(struct urb *);
  * The USB_ASYNC_UNLINK transfer flag affects later invocations of
  * the usb_unlink_urb() routine.
  *
- * All non-isochronous URBs must also initialize 
+ * All URBs must also initialize 
  * transfer_buffer and transfer_buffer_length.  They may provide the
  * URB_SHORT_NOT_OK transfer flag, indicating that short reads are
- * to be treated as errors.
+ * to be treated as errors; that flag is invalid for write requests.
  *
  * Bulk URBs may
  * use the USB_ZERO_PACKET transfer flag, indicating that bulk OUT transfers
@@ -896,10 +923,12 @@ struct urb
 	int status;			/* (return) non-ISO status */
 	unsigned int transfer_flags;	/* (in) URB_SHORT_NOT_OK | ...*/
 	void *transfer_buffer;		/* (in) associated data buffer */
+	dma_addr_t transfer_dma;	/* (in) dma addr for transfer_buffer */
 	int transfer_buffer_length;	/* (in) data buffer length */
 	int actual_length;		/* (return) actual transfer length */
 	int bandwidth;			/* bandwidth for INT/ISO request */
 	unsigned char *setup_packet;	/* (in) setup packet (control only) */
+	dma_addr_t setup_dma;		/* (in) dma addr for setup_packet */
 	int start_frame;		/* (modify) start frame (INT/ISO) */
 	int number_of_packets;		/* (in) number of ISO packets */
 	int interval;			/* (in) transfer interval (INT/ISO) */
@@ -910,6 +939,8 @@ struct urb
 	struct usb_iso_packet_descriptor iso_frame_desc[0];	/* (in) ISO ONLY */
 };
 
+/* -------------------------------------------------------------------------- */
+
 /**
  * usb_fill_control_urb - initializes a control urb
  * @urb: pointer to the urb to initialize.
@@ -1032,6 +1063,16 @@ extern struct urb *usb_get_urb(struct urb *urb);
 extern int usb_submit_urb(struct urb *urb, int mem_flags);
 extern int usb_unlink_urb(struct urb *urb);
 
+#define HAVE_USB_BUFFERS
+void *usb_buffer_alloc (struct usb_device *dev, size_t size,
+	int mem_flags, dma_addr_t *dma);
+void usb_buffer_free (struct usb_device *dev, size_t size,
+	void *addr, dma_addr_t dma);
+
+struct urb *usb_buffer_map (struct urb *urb);
+void usb_buffer_dmasync (struct urb *urb);
+void usb_buffer_unmap (struct urb *urb);
+
 /*-------------------------------------------------------------------*
  *                         SYNCHRONOUS CALL SUPPORT                  *
  *-------------------------------------------------------------------*/

include/linux/vmalloc.h

-#ifndef __LINUX_VMALLOC_H
-#define __LINUX_VMALLOC_H
+#ifndef _LINUX_VMALLOC_H
+#define _LINUX_VMALLOC_H
 
 #include <linux/spinlock.h>
 
-#include <asm/pgtable.h>
-
 /* bits in vm_struct->flags */
 #define VM_IOREMAP	0x00000001	/* ioremap() and friends */
 #define VM_ALLOC	0x00000002	/* vmalloc() */
+#define VM_MAP		0x00000004	/* vmap()ed pages */
 
 struct vm_struct {
-	unsigned long flags;
-	void * addr;
-	unsigned long size;
-	unsigned long phys_addr;
-	struct vm_struct * next;
+	void			*addr;
+	unsigned long		size;
+	unsigned long		flags;
+	struct page		**pages;
+	unsigned int		nr_pages;
+	unsigned long		phys_addr;
+	struct vm_struct	*next;
 };
 
-extern struct vm_struct * get_vm_area (unsigned long size, unsigned long flags);
-extern void vfree(void * addr);
-extern void * __vmalloc (unsigned long size, int gfp_mask, pgprot_t prot);
-extern long vread(char *buf, char *addr, unsigned long count);
-extern void vmfree_area_pages(unsigned long address, unsigned long size);
-extern int vmalloc_area_pages(unsigned long address, unsigned long size,
-                              int gfp_mask, pgprot_t prot);
-extern struct vm_struct *remove_kernel_area(void *addr);
-
 /*
- * Various ways to allocate pages.
+ *	Highlevel APIs for driver use
  */
-
-extern void * vmalloc(unsigned long size);
-extern void * vmalloc_32(unsigned long size);
+extern void *vmalloc(unsigned long size);
+extern void *vmalloc_32(unsigned long size);
+extern void *__vmalloc(unsigned long size, int gfp_mask, pgprot_t prot);
+extern void vfree(void *addr);
+
+extern void *vmap(struct page **pages, unsigned int count);
+extern void vunmap(void *addr);
+ 
+/*
+ *	Lowlevel-APIs (not for driver use!)
+ */
+extern struct vm_struct *get_vm_area(unsigned long size, unsigned long flags);
+extern struct vm_struct *remove_vm_area(void *addr);
+extern int map_vm_area(struct vm_struct *area, pgprot_t prot,
+			struct page ***pages);
+extern void unmap_vm_area(struct vm_struct *area);
 
 /*
- * vmlist_lock is a read-write spinlock that protects vmlist
- * Used in mm/vmalloc.c (get_vm_area() and vfree()) and fs/proc/kcore.c.
+ *	Internals.  Dont't use..
  */
 extern rwlock_t vmlist_lock;
+extern struct vm_struct *vmlist;
 
-extern struct vm_struct * vmlist;
-#endif
-
+#endif /* _LINUX_VMALLOC_H */

include/linux/wait.h

@@ -43,16 +43,16 @@ typedef struct __wait_queue_head wait_queue_head_t;
  */
 
 #define __WAITQUEUE_INITIALIZER(name, tsk) {				\
-	task:		tsk,						\
-	func:		default_wake_function,				\
-	task_list:	{ NULL, NULL } }
+	.task		= tsk,						\
+	.func		= default_wake_function,			\
+	.task_list	= { NULL, NULL } }
 
 #define DECLARE_WAITQUEUE(name, tsk)					\
 	wait_queue_t name = __WAITQUEUE_INITIALIZER(name, tsk)
 
 #define __WAIT_QUEUE_HEAD_INITIALIZER(name) {				\
-	lock:		SPIN_LOCK_UNLOCKED,				\
-	task_list:	{ &(name).task_list, &(name).task_list } }
+	.lock		= SPIN_LOCK_UNLOCKED,				\
+	.task_list	= { &(name).task_list, &(name).task_list } }
 
 #define DECLARE_WAIT_QUEUE_HEAD(name) \
 	wait_queue_head_t name = __WAIT_QUEUE_HEAD_INITIALIZER(name)

init/do_mounts.c

@@ -24,16 +24,16 @@
 
 extern int get_filesystem_list(char * buf);
 
-asmlinkage long sys_mount(char *dev_name, char *dir_name, char *type,
+extern asmlinkage long sys_mount(char *dev_name, char *dir_name, char *type,
 	 unsigned long flags, void *data);
-asmlinkage long sys_mkdir(char *name, int mode);
-asmlinkage long sys_chdir(char *name);
-asmlinkage long sys_chroot(char *name);
-asmlinkage long sys_unlink(char *name);
-asmlinkage long sys_symlink(char *old, char *new);
-asmlinkage long sys_mknod(char *name, int mode, dev_t dev);
-asmlinkage long sys_umount(char *name, int flags);
-asmlinkage long sys_ioctl(int fd, int cmd, unsigned long arg);
+extern asmlinkage long sys_mkdir(const char *name, int mode);
+extern asmlinkage long sys_chdir(const char *name);
+extern asmlinkage long sys_chroot(const char *name);
+extern asmlinkage long sys_unlink(const char *name);
+extern asmlinkage long sys_symlink(const char *old, const char *new);
+extern asmlinkage long sys_mknod(const char *name, int mode, dev_t dev);
+extern asmlinkage long sys_umount(char *name, int flags);
+extern asmlinkage long sys_ioctl(int fd, int cmd, unsigned long arg);
 
 #ifdef CONFIG_BLK_DEV_INITRD
 unsigned int real_root_dev;	/* do_proc_dointvec cannot handle kdev_t */

kernel/ksyms.c

@@ -109,6 +109,8 @@ EXPORT_SYMBOL(vfree);
 EXPORT_SYMBOL(__vmalloc);
 EXPORT_SYMBOL(vmalloc);
 EXPORT_SYMBOL(vmalloc_32);
+EXPORT_SYMBOL(vmap);
+EXPORT_SYMBOL(vunmap);
 EXPORT_SYMBOL(vmalloc_to_page);
 EXPORT_SYMBOL(mem_map);
 EXPORT_SYMBOL(remap_page_range);
@@ -220,6 +222,7 @@ EXPORT_SYMBOL(generic_file_read);
 EXPORT_SYMBOL(generic_file_sendfile);
 EXPORT_SYMBOL(do_generic_file_read);
 EXPORT_SYMBOL(generic_file_write);
+EXPORT_SYMBOL(generic_file_write_nolock);
 EXPORT_SYMBOL(generic_file_mmap);
 EXPORT_SYMBOL(generic_ro_fops);
 EXPORT_SYMBOL(file_lock_list);

mm/filemap.c

@@ -1109,7 +1109,7 @@ static int file_send_actor(read_descriptor_t * desc, struct page *page, unsigned
 	return written;
 }
 
-ssize_t generic_file_sendfile(struct file *in_file, struct file *out_file,
+ssize_t generic_file_sendfile(struct file *out_file, struct file *in_file,
 			      loff_t *ppos, size_t count)
 {
 	read_descriptor_t desc;
@@ -1812,9 +1812,8 @@ inline void remove_suid(struct dentry *dentry)
  * it for writing by marking it dirty.
  *							okir@monad.swb.de
  */
-ssize_t
-generic_file_write(struct file *file, const char *buf,
-		size_t count, loff_t *ppos)
+ssize_t generic_file_write_nolock(struct file *file, const char *buf,
+				  size_t count, loff_t *ppos)
 {
 	struct address_space * mapping = file->f_dentry->d_inode->i_mapping;
 	struct address_space_operations *a_ops = mapping->a_ops;
@@ -1829,18 +1828,15 @@ generic_file_write(struct file *file, const char *buf,
 	unsigned	bytes;
 	time_t		time_now;
 
-	if (unlikely((ssize_t) count < 0))
+	if (unlikely((ssize_t)count < 0))
 		return -EINVAL;
 
 	if (unlikely(!access_ok(VERIFY_READ, buf, count)))
 		return -EFAULT;
 
-	down(&inode->i_sem);
 	pos = *ppos;
-	if (unlikely(pos < 0)) {
-		err = -EINVAL;
-		goto out;
-	}
+	if (unlikely(pos < 0))
+		return -EINVAL;
 
 	if (unlikely(file->f_error)) {
 		err = file->f_error;
@@ -2037,6 +2033,18 @@ generic_file_write(struct file *file, const char *buf,
 out_status:	
 	err = written ? written : status;
 out:
+	return err;
+}
+
+ssize_t generic_file_write(struct file *file, const char *buf,
+			   size_t count, loff_t *ppos)
+{
+	struct inode	*inode = file->f_dentry->d_inode->i_mapping->host;
+	int		err;
+
+	down(&inode->i_sem);
+	err = generic_file_write_nolock(file, buf, count, ppos);
 	up(&inode->i_sem);
+
 	return err;
 }

mm/vmalloc.c

@@ -4,27 +4,28 @@
  *  Copyright (C) 1993  Linus Torvalds
  *  Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
  *  SMP-safe vmalloc/vfree/ioremap, Tigran Aivazian <tigran@veritas.com>, May 2000
+ *  Major rework to support vmap/vunmap, Christoph Hellwig, SGI, August 2002
  */
 
-#include <linux/config.h>
-#include <linux/slab.h>
-#include <linux/vmalloc.h>
-#include <linux/spinlock.h>
 #include <linux/mm.h>
 #include <linux/highmem.h>
-#include <linux/smp_lock.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/vmalloc.h>
 
 #include <asm/uaccess.h>
 #include <asm/pgalloc.h>
 #include <asm/tlbflush.h>
 
+
 rwlock_t vmlist_lock = RW_LOCK_UNLOCKED;
-struct vm_struct * vmlist;
+struct vm_struct *vmlist;
 
-static inline void free_area_pte(pmd_t * pmd, unsigned long address, unsigned long size)
+static inline void unmap_area_pte(pmd_t *pmd, unsigned long address,
+				  unsigned long size)
 {
-	pte_t * pte;
 	unsigned long end;
+	pte_t *pte;
 
 	if (pmd_none(*pmd))
 		return;
@@ -33,11 +34,13 @@ static inline void free_area_pte(pmd_t * pmd, unsigned long address, unsigned lo
 		pmd_clear(pmd);
 		return;
 	}
+
 	pte = pte_offset_kernel(pmd, address);
 	address &= ~PMD_MASK;
 	end = address + size;
 	if (end > PMD_SIZE)
 		end = PMD_SIZE;
+
 	do {
 		pte_t page;
 		page = ptep_get_and_clear(pte);
@@ -45,24 +48,17 @@ static inline void free_area_pte(pmd_t * pmd, unsigned long address, unsigned lo
 		pte++;
 		if (pte_none(page))
 			continue;
-		if (pte_present(page)) {
-			struct page *ptpage;
-			unsigned long pfn = pte_pfn(page);
-			if (!pfn_valid(pfn))
-				continue;
-			ptpage = pfn_to_page(pfn);
-			if (!PageReserved(ptpage))
-				__free_page(ptpage);
+		if (pte_present(page))
 			continue;
-		}
 		printk(KERN_CRIT "Whee.. Swapped out page in kernel page table\n");
 	} while (address < end);
 }
 
-static inline void free_area_pmd(pgd_t * dir, unsigned long address, unsigned long size)
+static inline void unmap_area_pmd(pgd_t *dir, unsigned long address,
+				  unsigned long size)
 {
-	pmd_t * pmd;
 	unsigned long end;
+	pmd_t *pmd;
 
 	if (pgd_none(*dir))
 		return;
@@ -71,36 +67,23 @@ static inline void free_area_pmd(pgd_t * dir, unsigned long address, unsigned lo
 		pgd_clear(dir);
 		return;
 	}
+
 	pmd = pmd_offset(dir, address);
 	address &= ~PGDIR_MASK;
 	end = address + size;
 	if (end > PGDIR_SIZE)
 		end = PGDIR_SIZE;
+
 	do {
-		free_area_pte(pmd, address, end - address);
+		unmap_area_pte(pmd, address, end - address);
 		address = (address + PMD_SIZE) & PMD_MASK;
 		pmd++;
 	} while (address < end);
 }
 
-void vmfree_area_pages(unsigned long start, unsigned long size)
-{
-	pgd_t * dir;
-	unsigned long address = start;
-	unsigned long end = start + size;
-
-	dir = pgd_offset_k(address);
-	flush_cache_all();
-	do {
-		free_area_pmd(dir, address, end - address);
-		address = (address + PGDIR_SIZE) & PGDIR_MASK;
-		dir++;
-	} while (address && (address < end));
-	flush_tlb_kernel_range(start, end);
-}
-
-static inline int alloc_area_pte (pte_t * pte, unsigned long address,
-			unsigned long size, int gfp_mask, pgprot_t prot)
+static inline int map_area_pte(pte_t *pte, unsigned long address,
+			       unsigned long size, pgprot_t prot,
+			       struct page ***pages)
 {
 	unsigned long end;
 
@@ -108,23 +91,26 @@ static inline int alloc_area_pte (pte_t * pte, unsigned long address,
 	end = address + size;
 	if (end > PMD_SIZE)
 		end = PMD_SIZE;
+
 	do {
-		struct page * page;
-		spin_unlock(&init_mm.page_table_lock);
-		page = alloc_page(gfp_mask);
-		spin_lock(&init_mm.page_table_lock);
+		struct page *page = **pages;
+
 		if (!pte_none(*pte))
 			printk(KERN_ERR "alloc_area_pte: page already exists\n");
 		if (!page)
 			return -ENOMEM;
+
 		set_pte(pte, mk_pte(page, prot));
 		address += PAGE_SIZE;
 		pte++;
+		(*pages)++;
 	} while (address < end);
 	return 0;
 }
 
-static inline int alloc_area_pmd(pmd_t * pmd, unsigned long address, unsigned long size, int gfp_mask, pgprot_t prot)
+static inline int map_area_pmd(pmd_t *pmd, unsigned long address,
+			       unsigned long size, pgprot_t prot,
+			       struct page ***pages)
 {
 	unsigned long end;
 
@@ -132,76 +118,108 @@ static inline int alloc_area_pmd(pmd_t * pmd, unsigned long address, unsigned lo
 	end = address + size;
 	if (end > PGDIR_SIZE)
 		end = PGDIR_SIZE;
+
 	do {
 		pte_t * pte = pte_alloc_kernel(&init_mm, pmd, address);
 		if (!pte)
 			return -ENOMEM;
-		if (alloc_area_pte(pte, address, end - address, gfp_mask, prot))
+		if (map_area_pte(pte, address, end - address, prot, pages))
 			return -ENOMEM;
 		address = (address + PMD_SIZE) & PMD_MASK;
 		pmd++;
 	} while (address < end);
+
 	return 0;
 }
 
-inline int vmalloc_area_pages (unsigned long address, unsigned long size,
-                               int gfp_mask, pgprot_t prot)
+void unmap_vm_area(struct vm_struct *area)
+{
+	unsigned long address = VMALLOC_VMADDR(area->addr);
+	unsigned long end = (address + area->size);
+	pgd_t *dir;
+
+	dir = pgd_offset_k(address);
+	flush_cache_all();
+	do {
+		unmap_area_pmd(dir, address, end - address);
+		address = (address + PGDIR_SIZE) & PGDIR_MASK;
+		dir++;
+	} while (address && (address < end));
+	flush_tlb_kernel_range(VMALLOC_VMADDR(area->addr), end);
+}
+
+int map_vm_area(struct vm_struct *area, pgprot_t prot, struct page ***pages)
 {
-	pgd_t * dir;
-	unsigned long end = address + size;
-	int ret;
+	unsigned long address = VMALLOC_VMADDR(area->addr);
+	unsigned long end = address + (area->size-PAGE_SIZE);
+	pgd_t *dir;
 
 	dir = pgd_offset_k(address);
 	spin_lock(&init_mm.page_table_lock);
 	do {
-		pmd_t *pmd;
-		
-		pmd = pmd_alloc(&init_mm, dir, address);
-		ret = -ENOMEM;
+		pmd_t *pmd = pmd_alloc(&init_mm, dir, address);
 		if (!pmd)
-			break;
-
-		ret = -ENOMEM;
-		if (alloc_area_pmd(pmd, address, end - address, gfp_mask, prot))
-			break;
+			return -ENOMEM;
+		if (map_area_pmd(pmd, address, end - address, prot, pages))
+			return -ENOMEM;
 
 		address = (address + PGDIR_SIZE) & PGDIR_MASK;
 		dir++;
-
-		ret = 0;
 	} while (address && (address < end));
+
 	spin_unlock(&init_mm.page_table_lock);
 	flush_cache_all();
-	return ret;
+	return 0;
 }
 
-struct vm_struct * get_vm_area(unsigned long size, unsigned long flags)
+
+/**
+ *	get_vm_area  -  reserve a contingous kernel virtual area
+ *
+ *	@size:		size of the area
+ *	@flags:		%VM_IOREMAP for I/O mappings or VM_ALLOC
+ *
+ *	Search an area of @size in the kernel virtual mapping area,
+ *	and reserved it for out purposes.  Returns the area descriptor
+ *	on success or %NULL on failure.
+ */
+struct vm_struct *get_vm_area(unsigned long size, unsigned long flags)
 {
-	unsigned long addr;
 	struct vm_struct **p, *tmp, *area;
+	unsigned long addr = VMALLOC_START;
 
-	area = (struct vm_struct *) kmalloc(sizeof(*area), GFP_KERNEL);
-	if (!area)
+	area = kmalloc(sizeof(*area), GFP_KERNEL);
+	if (unlikely(!area))
 		return NULL;
+
+	/*
+	 * We always allocate a guard page.
+	 */
 	size += PAGE_SIZE;
-	addr = VMALLOC_START;
+
 	write_lock(&vmlist_lock);
-	for (p = &vmlist; (tmp = *p) ; p = &tmp->next) {
+	for (p = &vmlist; (tmp = *p) ;p = &tmp->next) {
 		if ((size + addr) < addr)
 			goto out;
-		if (size + addr <= (unsigned long) tmp->addr)
-			break;
-		addr = tmp->size + (unsigned long) tmp->addr;
+		if (size + addr <= (unsigned long)tmp->addr)
+			goto found;
+		addr = tmp->size + (unsigned long)tmp->addr;
 		if (addr > VMALLOC_END-size)
 			goto out;
 	}
-	area->phys_addr = 0;
+
+found:
+	area->next = *p;
+	*p = area;
+
 	area->flags = flags;
 	area->addr = (void *)addr;
 	area->size = size;
-	area->next = *p;
-	*p = area;
+	area->pages = NULL;
+	area->nr_pages = 0;
+	area->phys_addr = 0;
 	write_unlock(&vmlist_lock);
+
 	return area;
 
 out:
@@ -210,87 +228,203 @@ struct vm_struct * get_vm_area(unsigned long size, unsigned long flags)
 	return NULL;
 }
 
-struct vm_struct *remove_kernel_area(void *addr) 
+/**
+ *	remove_vm_area  -  find and remove a contingous kernel virtual area
+ *
+ *	@addr:		base address
+ *
+ *	Search for the kernel VM area starting at @addr, and remove it.
+ *	This function returns the found VM area, but using it is NOT safe
+ *	on SMP machines.
+ */
+struct vm_struct *remove_vm_area(void *addr)
 {
 	struct vm_struct **p, *tmp;
-	write_lock(&vmlist_lock);
-	for (p = &vmlist ; (tmp = *p) ; p = &tmp->next) {
-		if (tmp->addr == addr) {
-			*p = tmp->next;
-			write_unlock(&vmlist_lock);
-			return tmp;
-		}
 
+	write_lock(&vmlist_lock);
+	for (p = &vmlist ; (tmp = *p) ;p = &tmp->next) {
+		 if (tmp->addr == addr)
+			 goto found;
 	}
 	write_unlock(&vmlist_lock);
 	return NULL;
-} 
 
-void vfree(void * addr)
+found:
+	*p = tmp->next;
+	write_unlock(&vmlist_lock);
+	return tmp;
+}
+
+void __vunmap(void *addr, int deallocate_pages)
 {
-	struct vm_struct *tmp;
+	struct vm_struct *area;
 
 	if (!addr)
 		return;
-	if ((PAGE_SIZE-1) & (unsigned long) addr) {
+
+	if ((PAGE_SIZE-1) & (unsigned long)addr) {
 		printk(KERN_ERR "Trying to vfree() bad address (%p)\n", addr);
 		return;
 	}
-	tmp = remove_kernel_area(addr); 
-	if (tmp) { 
-			vmfree_area_pages(VMALLOC_VMADDR(tmp->addr), tmp->size);
-			kfree(tmp);
-			return;
+
+	area = remove_vm_area(addr);
+	if (unlikely(!area)) {
+		printk(KERN_ERR "Trying to vfree() nonexistent vm area (%p)\n",
+				addr);
+		return;
+	}
+
+	unmap_vm_area(area);
+	
+	if (deallocate_pages) {
+		int i;
+
+		for (i = 0; i < area->nr_pages; i++) {
+			if (unlikely(!area->pages[i]))
+				BUG();
+			__free_page(area->pages[i]);
 		}
-	printk(KERN_ERR "Trying to vfree() nonexistent vm area (%p)\n", addr);
+
+		kfree(area->pages);
+	}
+
+	kfree(area);
+	return;
 }
 
-/*
- *	Allocate any pages
+/**
+ *	vfree  -  release memory allocated by vmalloc()
+ *
+ *	@addr:		memory base address
+ *
+ *	Free the virtually continguos memory area starting at @addr, as
+ *	obtained from vmalloc(), vmalloc_32() or __vmalloc().
+ *
+ *	May not be called in interrupt context.
  */
-
-void * vmalloc (unsigned long size)
+void vfree(void *addr)
 {
-	return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM, PAGE_KERNEL);
+	__vunmap(addr, 1);
 }
 
-/*
- *	Allocate ISA addressable pages for broke crap
+/**
+ *	vunmap  -  release virtual mapping obtained by vmap()
+ *
+ *	@addr:		memory base address
+ *
+ *	Free the virtually continguos memory area starting at @addr,
+ *	which was created from the page array passed to vmap().
+ *
+ *	May not be called in interrupt context.
  */
-
-void * vmalloc_dma (unsigned long size)
+void vunmap(void *addr)
 {
-	return __vmalloc(size, GFP_KERNEL|GFP_DMA, PAGE_KERNEL);
+	__vunmap(addr, 0);
 }
 
-/*
- *	vmalloc 32bit PA addressable pages - eg for PCI 32bit devices
+/**
+ *	vmap  -  map an array of pages into virtually continguos space
+ *
+ *	@pages:		array of page pointers
+ *	@count:		number of pages to map
+ *
+ *	Maps @count pages from @pages into continguos kernel virtual
+ *	space.
  */
-
-void * vmalloc_32(unsigned long size)
+void *vmap(struct page **pages, unsigned int count)
 {
-	return __vmalloc(size, GFP_KERNEL, PAGE_KERNEL);
+	struct vm_struct *area;
+
+	if (count > num_physpages)
+		return NULL;
+
+	area = get_vm_area((count << PAGE_SHIFT), VM_MAP);
+	if (!area)
+		return NULL;
+	if (map_vm_area(area, PAGE_KERNEL, &pages)) {
+		vunmap(area->addr);
+		return NULL;
+	}
+
+	return area->addr;
 }
 
-void * __vmalloc (unsigned long size, int gfp_mask, pgprot_t prot)
+/**
+ *	__vmalloc  -  allocate virtually continguos memory
+ *
+ *	@size:		allocation size
+ *	@gfp_mask:	flags for the page level allocator
+ *	@prot:		protection mask for the allocated pages
+ *
+ *	Allocate enough pages to cover @size from the page level
+ *	allocator with @gfp_mask flags.  Map them into continguos
+ *	kernel virtual space, using a pagetable protection of @prot.
+ */
+void *__vmalloc(unsigned long size, int gfp_mask, pgprot_t prot)
 {
-	void * addr;
 	struct vm_struct *area;
+	struct page **pages;
+	unsigned int nr_pages, array_size, i;
 
 	size = PAGE_ALIGN(size);
-	if (!size || (size >> PAGE_SHIFT) > num_physpages) {
-		BUG();
+	if (!size || (size >> PAGE_SHIFT) > num_physpages)
 		return NULL;
-	}
+
 	area = get_vm_area(size, VM_ALLOC);
 	if (!area)
 		return NULL;
-	addr = area->addr;
-	if (vmalloc_area_pages(VMALLOC_VMADDR(addr), size, gfp_mask, prot)) {
-		vfree(addr);
+
+	nr_pages = (size+PAGE_SIZE) >> PAGE_SHIFT;
+	array_size = (nr_pages * sizeof(struct page *));
+
+	area->nr_pages = nr_pages;
+	area->pages = pages = kmalloc(array_size, (gfp_mask & ~__GFP_HIGHMEM));
+	if (!area->pages)
 		return NULL;
+	memset(area->pages, 0, array_size);
+
+	for (i = 0; i < area->nr_pages; i++) {
+		area->pages[i] = alloc_page(gfp_mask);
+		if (unlikely(!area->pages[i]))
+			goto fail;
 	}
-	return addr;
+	
+	if (map_vm_area(area, prot, &pages))
+		goto fail;
+	return area->addr;
+
+fail:
+	vfree(area->addr);
+	return NULL;
+}
+
+/**
+ *	vmalloc  -  allocate virtually continguos memory
+ *
+ *	@size:		allocation size
+ *
+ *	Allocate enough pages to cover @size from the page level
+ *	allocator and map them into continguos kernel virtual space.
+ *
+ *	For tight cotrol over page level allocator and protection flags
+ *	use __vmalloc() instead.
+ */
+void *vmalloc(unsigned long size)
+{
+       return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM, PAGE_KERNEL);
+}
+
+/**
+ *	vmalloc_32  -  allocate virtually continguos memory (32bit addressable)
+ *
+ *	@size:		allocation size
+ *
+ *	Allocate enough 32bit PA addressable pages to cover @size from the
+ *	page level allocator and map them into continguos kernel virtual space.
+ */
+void *vmalloc_32(unsigned long size)
+{
+	return __vmalloc(size, GFP_KERNEL, PAGE_KERNEL);
 }
 
 long vread(char *buf, char *addr, unsigned long count)

net/sunrpc/svcsock.c

@@ -481,6 +481,16 @@ svc_udp_recvfrom(struct svc_rqst *rqstp)
 	u32		*data;
 	int		err, len;
 
+	if (test_and_clear_bit(SK_CHNGBUF, &svsk->sk_flags))
+	    /* udp sockets need large rcvbuf as all pending
+	     * requests are still in that buffer.  sndbuf must
+	     * also be large enough that there is enough space
+	     * for one reply per thread.
+	     */
+	    svc_sock_setbufsize(svsk->sk_sock,
+				(serv->sv_nrthreads+3) * serv->sv_bufsz,
+				(serv->sv_nrthreads+3) * serv->sv_bufsz);
+
 	clear_bit(SK_DATA, &svsk->sk_flags);
 	while ((skb = skb_recv_datagram(svsk->sk_sk, 0, 1, &err)) == NULL) {
 		svc_sock_received(svsk);
@@ -564,6 +574,8 @@ svc_udp_init(struct svc_sock *svsk)
 	svsk->sk_recvfrom = svc_udp_recvfrom;
 	svsk->sk_sendto = svc_udp_sendto;
 
+	set_bit(SK_CHNGBUF, &svsk->sk_flags);
+
 	return 0;
 }
 
@@ -771,6 +783,18 @@ svc_tcp_recvfrom(struct svc_rqst *rqstp)
 		return 0;
 	}
 
+	if (test_and_clear_bit(SK_CHNGBUF, &svsk->sk_flags))
+		/* sndbuf needs to have room for one request
+		 * per thread, otherwise we can stall even when the
+		 * network isn't a bottleneck.
+		 * rcvbuf just needs to be able to hold a few requests.
+		 * Normally they will be removed from the queue 
+		 * as soon a a complete request arrives.
+		 */
+		svc_sock_setbufsize(svsk->sk_sock,
+				    (serv->sv_nrthreads+3) * serv->sv_bufsz,
+				    3 * serv->sv_bufsz);
+
 	clear_bit(SK_DATA, &svsk->sk_flags);
 
 	/* Receive data. If we haven't got the record length yet, get
@@ -787,6 +811,9 @@ svc_tcp_recvfrom(struct svc_rqst *rqstp)
 			goto error;
 		svsk->sk_tcplen += len;
 
+		if (len < want)
+			return 0;
+
 		svsk->sk_reclen = ntohl(svsk->sk_reclen);
 		if (!(svsk->sk_reclen & 0x80000000)) {
 			/* FIXME: technically, a record can be fragmented,
@@ -887,8 +914,9 @@ svc_tcp_sendto(struct svc_rqst *rqstp)
 
 	sent = svc_sendto(rqstp, bufp->iov, bufp->nriov);
 	if (sent != bufp->len<<2) {
-		printk(KERN_NOTICE "rpc-srv/tcp: %s: sent only %d bytes of %d - should shutdown socket\n",
+		printk(KERN_NOTICE "rpc-srv/tcp: %s: %s %d when sending %d bytes - shutting down socket\n",
 		       rqstp->rq_sock->sk_server->sv_name,
+		       (sent<0)?"got error":"sent only",
 		       sent, bufp->len << 2);
 		svc_delete_socket(rqstp->rq_sock);
 		sent = -EAGAIN;
@@ -916,17 +944,7 @@ svc_tcp_init(struct svc_sock *svsk)
 		svsk->sk_reclen = 0;
 		svsk->sk_tcplen = 0;
 
-		/* sndbuf needs to have room for one request
-		 * per thread, otherwise we can stall even when the
-		 * network isn't a bottleneck.
-		 * rcvbuf just needs to be able to hold a few requests.
-		 * Normally they will be removed from the queue 
-		 * as soon a a complete request arrives.
-		 */
-		svc_sock_setbufsize(svsk->sk_sock,
-				    (svsk->sk_server->sv_nrthreads+3) *
-				    svsk->sk_server->sv_bufsz,
-				    3 * svsk->sk_server->sv_bufsz);
+		set_bit(SK_CHNGBUF, &svsk->sk_flags);
 	}
 
 	return 0;
@@ -945,30 +963,12 @@ svc_sock_update_bufs(struct svc_serv *serv)
 	list_for_each(le, &serv->sv_permsocks) {
 		struct svc_sock *svsk = 
 			list_entry(le, struct svc_sock, sk_list);
-		struct socket *sock = svsk->sk_sock;
-		if (sock->type == SOCK_DGRAM) {
-			/* udp sockets need large rcvbuf as all pending
-			 * requests are still in that buffer.
-			 */
-			svc_sock_setbufsize(sock,
-					    (serv->sv_nrthreads+3) * serv->sv_bufsz,
-					    (serv->sv_nrthreads+3) * serv->sv_bufsz);
-		} else if (svsk->sk_sk->state != TCP_LISTEN) {
-			printk(KERN_ERR "RPC update_bufs: permanent sock neither UDP or TCP_LISTEN\n");
-		}
+		set_bit(SK_CHNGBUF, &svsk->sk_flags);
 	}
 	list_for_each(le, &serv->sv_tempsocks) {
 		struct svc_sock *svsk =
 			list_entry(le, struct svc_sock, sk_list);
-		struct socket *sock = svsk->sk_sock;
-		if (sock->type == SOCK_STREAM) {
-			/* See svc_tcp_init above for rationale on buffer sizes */
-			svc_sock_setbufsize(sock,
-					    (serv->sv_nrthreads+3) *
-					    serv->sv_bufsz,
-					    3 * serv->sv_bufsz);
-		} else 
-			printk(KERN_ERR "RPC update_bufs: temp sock not TCP\n");
+		set_bit(SK_CHNGBUF, &svsk->sk_flags);
 	}
 	spin_unlock_bh(&serv->sv_lock);
 }
@@ -1212,6 +1212,7 @@ svc_create_socket(struct svc_serv *serv, int protocol, struct sockaddr_in *sin)
 		return error;
 
 	if (sin != NULL) {
+		sock->sk->reuse = 1; /* allow address reuse */
 		error = sock->ops->bind(sock, (struct sockaddr *) sin,
 						sizeof(*sin));
 		if (error < 0)