s390/mem_detect: move tprot loop to early boot phase

Move memory detection to early boot phase. To store online memory
regions "struct mem_detect_info" has been introduced together with
for_each_mem_detect_block iterator. mem_detect_info is later converted
to memblock.

Also introduces sclp_early_get_meminfo function to get maximum physical
memory and maximum increment number.
Reviewed-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Reviewed-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/boot/Makefile

@@ -27,7 +27,7 @@ endif
 
 CFLAGS_sclp_early_core.o += -I$(srctree)/drivers/s390/char
 
-obj-y	:= head.o als.o startup.o ebcdic.o sclp_early_core.o mem.o
+obj-y	:= head.o als.o startup.o mem_detect.o ebcdic.o sclp_early_core.o mem.o
 targets	:= bzImage startup.a $(obj-y)
 subdir-	:= compressed
 

arch/s390/boot/boot.h

@@ -3,5 +3,6 @@
 #define BOOT_BOOT_H
 
 void startup_kernel(void);
+void detect_memory(void);
 
 #endif /* BOOT_BOOT_H */

arch/s390/boot/mem_detect.c

+// SPDX-License-Identifier: GPL-2.0
+#include <linux/init.h>
+#include <asm/sclp.h>
+#include <asm/sections.h>
+#include <asm/mem_detect.h>
+#include "compressed/decompressor.h"
+#include "boot.h"
+
+#define CHUNK_READ_WRITE 0
+#define CHUNK_READ_ONLY  1
+
+unsigned long __bootdata(max_physmem_end);
+struct mem_detect_info __bootdata(mem_detect);
+
+/* up to 256 storage elements, 1020 subincrements each */
+#define ENTRIES_EXTENDED_MAX						       \
+	(256 * (1020 / 2) * sizeof(struct mem_detect_block))
+
+/*
+ * To avoid corrupting old kernel memory during dump, find lowest memory
+ * chunk possible either right after the kernel end (decompressed kernel) or
+ * after initrd (if it is present and there is no hole between the kernel end
+ * and initrd)
+ */
+static void *mem_detect_alloc_extended(void)
+{
+	unsigned long offset = ALIGN(mem_safe_offset(), sizeof(u64));
+
+	if (IS_ENABLED(BLK_DEV_INITRD) && INITRD_START && INITRD_SIZE &&
+	    INITRD_START < offset + ENTRIES_EXTENDED_MAX)
+		offset = ALIGN(INITRD_START + INITRD_SIZE, sizeof(u64));
+
+	return (void *)offset;
+}
+
+static struct mem_detect_block *__get_mem_detect_block_ptr(u32 n)
+{
+	if (n < MEM_INLINED_ENTRIES)
+		return &mem_detect.entries[n];
+	if (unlikely(!mem_detect.entries_extended))
+		mem_detect.entries_extended = mem_detect_alloc_extended();
+	return &mem_detect.entries_extended[n - MEM_INLINED_ENTRIES];
+}
+
+/*
+ * sequential calls to add_mem_detect_block with adjacent memory areas
+ * are merged together into single memory block.
+ */
+void add_mem_detect_block(u64 start, u64 end)
+{
+	struct mem_detect_block *block;
+
+	if (mem_detect.count) {
+		block = __get_mem_detect_block_ptr(mem_detect.count - 1);
+		if (block->end == start) {
+			block->end = end;
+			return;
+		}
+	}
+
+	block = __get_mem_detect_block_ptr(mem_detect.count);
+	block->start = start;
+	block->end = end;
+	mem_detect.count++;
+}
+
+static unsigned long get_mem_detect_end(void)
+{
+	if (mem_detect.count)
+		return __get_mem_detect_block_ptr(mem_detect.count - 1)->end;
+	return 0;
+}
+
+static int tprot(unsigned long addr)
+{
+	unsigned long pgm_addr;
+	int rc = -EFAULT;
+	psw_t old = S390_lowcore.program_new_psw;
+
+	S390_lowcore.program_new_psw.mask = __extract_psw();
+	asm volatile(
+		"	larl	%[pgm_addr],1f\n"
+		"	stg	%[pgm_addr],%[psw_pgm_addr]\n"
+		"	tprot	0(%[addr]),0\n"
+		"	ipm	%[rc]\n"
+		"	srl	%[rc],28\n"
+		"1:\n"
+		: [pgm_addr] "=&d"(pgm_addr),
+		  [psw_pgm_addr] "=Q"(S390_lowcore.program_new_psw.addr),
+		  [rc] "+&d"(rc)
+		: [addr] "a"(addr)
+		: "cc", "memory");
+	S390_lowcore.program_new_psw = old;
+	return rc;
+}
+
+static void scan_memory(unsigned long rzm)
+{
+	unsigned long addr, size;
+	int type;
+
+	if (!rzm)
+		rzm = 1UL << 20;
+
+	addr = 0;
+	do {
+		size = 0;
+		/* assume lowcore is writable */
+		type = addr ? tprot(addr) : CHUNK_READ_WRITE;
+		do {
+			size += rzm;
+			if (max_physmem_end && addr + size >= max_physmem_end)
+				break;
+		} while (type == tprot(addr + size));
+		if (type == CHUNK_READ_WRITE || type == CHUNK_READ_ONLY) {
+			if (max_physmem_end && (addr + size > max_physmem_end))
+				size = max_physmem_end - addr;
+			add_mem_detect_block(addr, addr + size);
+		}
+		addr += size;
+	} while (addr < max_physmem_end);
+}
+
+void detect_memory(void)
+{
+	unsigned long rzm;
+
+	sclp_early_get_meminfo(&max_physmem_end, &rzm);
+	scan_memory(rzm);
+	mem_detect.info_source = MEM_DETECT_TPROT_LOOP;
+	if (!max_physmem_end)
+		max_physmem_end = get_mem_detect_end();
+}

arch/s390/boot/startup.c

@@ -51,6 +51,7 @@ void startup_kernel(void)
 
 	rescue_initrd();
 	sclp_early_read_info();
+	detect_memory();
 	if (!IS_ENABLED(CONFIG_KERNEL_UNCOMPRESSED)) {
 		img = decompress_kernel();
 		memmove((void *)vmlinux.default_lma, img, vmlinux.image_size);

arch/s390/include/asm/mem_detect.h

+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_S390_MEM_DETECT_H
+#define _ASM_S390_MEM_DETECT_H
+
+#include <linux/types.h>
+
+enum mem_info_source {
+	MEM_DETECT_NONE = 0,
+	MEM_DETECT_TPROT_LOOP
+};
+
+struct mem_detect_block {
+	u64 start;
+	u64 end;
+};
+
+/*
+ * Storage element id is defined as 1 byte (up to 256 storage elements).
+ * In practise only storage element id 0 and 1 are used).
+ * According to architecture one storage element could have as much as
+ * 1020 subincrements. 255 mem_detect_blocks are embedded in mem_detect_info.
+ * If more mem_detect_blocks are required, a block of memory from already
+ * known mem_detect_block is taken (entries_extended points to it).
+ */
+#define MEM_INLINED_ENTRIES 255 /* (PAGE_SIZE - 16) / 16 */
+
+struct mem_detect_info {
+	u32 count;
+	u8 info_source;
+	struct mem_detect_block entries[MEM_INLINED_ENTRIES];
+	struct mem_detect_block *entries_extended;
+};
+extern struct mem_detect_info mem_detect;
+
+static inline int __get_mem_detect_block(u32 n, unsigned long *start,
+					 unsigned long *end)
+{
+	if (n >= mem_detect.count) {
+		*start = 0;
+		*end = 0;
+		return -1;
+	}
+
+	if (n < MEM_INLINED_ENTRIES) {
+		*start = (unsigned long)mem_detect.entries[n].start;
+		*end = (unsigned long)mem_detect.entries[n].end;
+	} else {
+		*start = (unsigned long)mem_detect.entries_extended[n - MEM_INLINED_ENTRIES].start;
+		*end = (unsigned long)mem_detect.entries_extended[n - MEM_INLINED_ENTRIES].end;
+	}
+	return 0;
+}
+
+/**
+ * for_each_mem_detect_block - early online memory range iterator
+ * @i: an integer used as loop variable
+ * @p_start: ptr to unsigned long for start address of the range
+ * @p_end: ptr to unsigned long for end address of the range
+ *
+ * Walks over detected online memory ranges.
+ */
+#define for_each_mem_detect_block(i, p_start, p_end)			\
+	for (i = 0, __get_mem_detect_block(i, p_start, p_end);		\
+	     i < mem_detect.count;					\
+	     i++, __get_mem_detect_block(i, p_start, p_end))
+
+static inline void get_mem_detect_reserved(unsigned long *start,
+					   unsigned long *size)
+{
+	*start = (unsigned long)mem_detect.entries_extended;
+	if (mem_detect.count > MEM_INLINED_ENTRIES)
+		*size = (mem_detect.count - MEM_INLINED_ENTRIES) * sizeof(struct mem_detect_block);
+	else
+		*size = 0;
+}
+
+#endif

arch/s390/include/asm/sclp.h

@@ -113,6 +113,7 @@ void sclp_early_printk(const char *s);
 void sclp_early_printk_force(const char *s);
 void __sclp_early_printk(const char *s, unsigned int len, unsigned int force);
 
+int sclp_early_get_meminfo(unsigned long *mem, unsigned long *rzm);
 int _sclp_get_core_info(struct sclp_core_info *info);
 int sclp_core_configure(u8 core);
 int sclp_core_deconfigure(u8 core);

arch/s390/include/asm/setup.h

@@ -69,8 +69,6 @@ extern int memory_end_set;
 extern unsigned long memory_end;
 extern unsigned long max_physmem_end;
 
-extern void detect_memory_memblock(void);
-
 #define MACHINE_IS_VM		(S390_lowcore.machine_flags & MACHINE_FLAG_VM)
 #define MACHINE_IS_KVM		(S390_lowcore.machine_flags & MACHINE_FLAG_KVM)
 #define MACHINE_IS_LPAR		(S390_lowcore.machine_flags & MACHINE_FLAG_LPAR)

arch/s390/kernel/setup.c

@@ -70,6 +70,7 @@
 #include <asm/numa.h>
 #include <asm/alternative.h>
 #include <asm/nospec-branch.h>
+#include <asm/mem_detect.h>
 #include "entry.h"
 
 /*
@@ -91,7 +92,8 @@ unsigned long int_hwcap = 0;
 
 int __initdata memory_end_set;
 unsigned long __initdata memory_end;
-unsigned long __initdata max_physmem_end;
+unsigned long __bootdata(max_physmem_end);
+struct mem_detect_info __bootdata(mem_detect);
 
 unsigned long VMALLOC_START;
 EXPORT_SYMBOL(VMALLOC_START);
@@ -720,6 +722,45 @@ static void __init reserve_initrd(void)
 #endif
 }
 
+static void __init reserve_mem_detect_info(void)
+{
+	unsigned long start, size;
+
+	get_mem_detect_reserved(&start, &size);
+	if (size)
+		memblock_reserve(start, size);
+}
+
+static void __init free_mem_detect_info(void)
+{
+	unsigned long start, size;
+
+	get_mem_detect_reserved(&start, &size);
+	if (size)
+		memblock_free(start, size);
+}
+
+static void __init memblock_physmem_add(phys_addr_t start, phys_addr_t size)
+{
+	memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n",
+		     start, start + size - 1);
+	memblock_add_range(&memblock.memory, start, size, 0, 0);
+	memblock_add_range(&memblock.physmem, start, size, 0, 0);
+}
+
+static void __init memblock_add_mem_detect_info(void)
+{
+	unsigned long start, end;
+	int i;
+
+	/* keep memblock lists close to the kernel */
+	memblock_set_bottom_up(true);
+	for_each_mem_detect_block(i, &start, &end)
+		memblock_physmem_add(start, end - start);
+	memblock_set_bottom_up(false);
+	memblock_dump_all();
+}
+
 /*
  * Check for initrd being in usable memory
  */
@@ -984,11 +1025,13 @@ void __init setup_arch(char **cmdline_p)
 	reserve_oldmem();
 	reserve_kernel();
 	reserve_initrd();
+	reserve_mem_detect_info();
 	memblock_allow_resize();
 
 	/* Get information about *all* installed memory */
-	detect_memory_memblock();
+	memblock_add_mem_detect_info();
 
+	free_mem_detect_info();
 	remove_oldmem();
 
 	/*

arch/s390/mm/Makefile

@@ -4,8 +4,7 @@
 #
 
 obj-y		:= init.o fault.o extmem.o mmap.o vmem.o maccess.o
-obj-y		+= page-states.o gup.o pageattr.o mem_detect.o
-obj-y		+= pgtable.o pgalloc.o
+obj-y		+= page-states.o gup.o pageattr.o pgtable.o pgalloc.o
 
 obj-$(CONFIG_CMM)		+= cmm.o
 obj-$(CONFIG_HUGETLB_PAGE)	+= hugetlbpage.o

arch/s390/mm/mem_detect.c

-// SPDX-License-Identifier: GPL-2.0
-/*
- * Copyright IBM Corp. 2008, 2009
- *
- * Author(s): Heiko Carstens <heiko.carstens@de.ibm.com>
- */
-
-#include <linux/kernel.h>
-#include <linux/memblock.h>
-#include <linux/init.h>
-#include <linux/debugfs.h>
-#include <linux/seq_file.h>
-#include <asm/ipl.h>
-#include <asm/sclp.h>
-#include <asm/setup.h>
-
-#define CHUNK_READ_WRITE 0
-#define CHUNK_READ_ONLY  1
-
-static inline void memblock_physmem_add(phys_addr_t start, phys_addr_t size)
-{
-	memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n",
-		     start, start + size - 1);
-	memblock_add_range(&memblock.memory, start, size, 0, 0);
-	memblock_add_range(&memblock.physmem, start, size, 0, 0);
-}
-
-void __init detect_memory_memblock(void)
-{
-	unsigned long memsize, rnmax, rzm, addr, size;
-	int type;
-
-	rzm = sclp.rzm;
-	rnmax = sclp.rnmax;
-	memsize = rzm * rnmax;
-	if (!rzm)
-		rzm = 1UL << 17;
-	max_physmem_end = memsize;
-	addr = 0;
-	/* keep memblock lists close to the kernel */
-	memblock_set_bottom_up(true);
-	do {
-		size = 0;
-		/* assume lowcore is writable */
-		type = addr ? tprot(addr) : CHUNK_READ_WRITE;
-		do {
-			size += rzm;
-			if (max_physmem_end && addr + size >= max_physmem_end)
-				break;
-		} while (type == tprot(addr + size));
-		if (type == CHUNK_READ_WRITE || type == CHUNK_READ_ONLY) {
-			if (max_physmem_end && (addr + size > max_physmem_end))
-				size = max_physmem_end - addr;
-			memblock_physmem_add(addr, size);
-		}
-		addr += size;
-	} while (addr < max_physmem_end);
-	memblock_set_bottom_up(false);
-	if (!max_physmem_end)
-		max_physmem_end = memblock_end_of_DRAM();
-	memblock_dump_all();
-}

drivers/s390/char/sclp_early_core.c

@@ -270,3 +270,20 @@ int __init sclp_early_get_info(struct read_info_sccb *info)
 	*info = sclp_info_sccb;
 	return 0;
 }
+
+int __init sclp_early_get_meminfo(unsigned long *mem, unsigned long *rzm)
+{
+	unsigned long rnmax;
+	unsigned long rnsize;
+	struct read_info_sccb *sccb = &sclp_info_sccb;
+
+	if (!sclp_info_sccb_valid)
+		return -EIO;
+
+	rnmax = sccb->rnmax ? sccb->rnmax : sccb->rnmax2;
+	rnsize = sccb->rnsize ? sccb->rnsize : sccb->rnsize2;
+	rnsize <<= 20;
+	*mem = rnsize * rnmax;
+	*rzm = rnsize;
+	return 0;
+}