- 30 Jan, 2008 40 commits
-
-
Mathieu Desnoyers authored
Actually, on 386, cmpxchg and cmpxchg_local fall back on cmpxchg_386_u8/16/32: it disables interruptions around non atomic updates to mimic the cmpxchg behavior. The comment: /* Poor man's cmpxchg for 386. Unsuitable for SMP */ already present in cmpxchg_386_u32 tells much about how this cmpxchg implementation should not be used in a SMP context. However, the cmpxchg_local can perfectly use this fallback, since it only needs to be atomic wrt the local cpu. This patch adds a cmpxchg_486_u64 and uses it as a fallback for cmpxchg64 and cmpxchg64_local on 80386 and 80486. Q: but why is it called cmpxchg_486 when the other functions are called A: Because the standard cmpxchg is missing only on 386, but cmpxchg8b is missing both on 386 and 486. Citing Intel's Instruction set reference: cmpxchg: This instruction is not supported on Intel processors earlier than the Intel486 processors. cmpxchg8b: This instruction encoding is not supported on Intel processors earlier than the Pentium processors. Q: What's the reason to have cmpxchg64_local on 32 bit architectures? Without that need all this would just be a few simple defines. A: cmpxchg64_local on 32 bits architectures takes unsigned long long parameters, but cmpxchg_local only takes longs. Since we have cmpxchg8b to execute a 8 byte cmpxchg atomically on pentium and +, it makes sense to provide a flavor of cmpxchg and cmpxchg_local using this instruction. Also, for 32 bits architectures lacking the 64 bits atomic cmpxchg, it makes sense _not_ to define cmpxchg64 while cmpxchg could still be available. Moreover, the fallback for cmpxchg8b on i386 for 386 and 486 is a However, cmpxchg64_local will be emulated by disabling interrupts on all architectures where it is not supported atomically. Therefore, we *could* turn cmpxchg64_local into a cmpxchg_local, but it would make the 386/486 fallbacks ugly, make its design different from cmpxchg/cmpxchg64 (which really depends on atomic operations and cannot be emulated) and require the __cmpxchg_local to be expressed as a macro rather than an inline function so the parameters would not be fixed to unsigned long long in every case. So I think cmpxchg64_local makes sense there, but I am open to suggestions. Q: Are there any callers? A: I am actually using it in LTTng in my timestamping code. I use it to work around CPUs with asynchronous TSCs. I need to update 64 bits values atomically on this 32 bits architecture. Changelog: - Ran though checkpatch. Signed-off-by:
Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca> Cc: Andi Kleen <ak@suse.de> Signed-off-by:
Andrew Morton <akpm@linux-foundation.org> Signed-off-by:
Ingo Molnar <mingo@elte.hu> Signed-off-by:
Thomas Gleixner <tglx@linutronix.de>
-
Ralf Baechle authored
The timer code always calls the clock_event_device set_net_event and set_mode methods with interrupts disabled, so no need to use spin_lock_irqsave / spin_unlock_irqrestore for those. Signed-off-by:
Ralf Baechle <ralf@linux-mips.org> Acked-by:
-
Christoph Lameter authored
Use sparsemem as the only memory model for UP, SMP and NUMA. Measurements indicate that DISCONTIGMEM has a higher overhead than sparsemem. And FLATMEMs benefits are minimal. So I think its best to simply standardize on sparsemem. Results of page allocator tests (test can be had via git from slab git tree branch tests) Measurements in cycle counts. 1000 allocations were performed and then the average cycle count was calculated. Order FlatMem Discontig SparseMem 0 639 665 641 1 567 647 593 2 679 774 692 3 763 967 781 4 961 1501 962 5 1356 2344 1392 6 2224 3982 2336 7 4869 7225 5074 8 12500 14048 12732 9 27926 28223 28165 10 58578 58714 58682 (Note that FlatMem is an SMP config and the rest NUMA configurations) Memory use: SMP Sparsemem ------------- Kernel size: text data bss dec hex filename 3849268 397739 1264856 5511863 541ab7 vmlinux total used free shared buffers cached Mem: 8242252 41164 8201088 0 352 11512 -/+ buffers/cache: 29300 8212952 Swap: 9775512 0 9775512 SMP Flatmem ----------- Kernel size: text data bss dec hex filename 3844612 397739 1264536 5506887 540747 vmlinux So 4.5k growth in text size vs. FLATMEM. total used free shared buffers cached Mem: 8244052 40544 8203508 0 352 11484 -/+ buffers/cache: 28708 8215344 2k growth in overall memory use after boot. NUMA discontig: text data bss dec hex filename 3888124 470659 1276504 5635287 55fcd7 vmlinux total used free shared buffers cached Mem: 8256256 56908 8199348 0 352 11496 -/+ buffers/cache: 45060 8211196 Swap: 9775512 0 9775512 NUMA sparse: text data bss dec hex filename 3896428 470659 1276824 5643911 561e87 vmlinux 8k text growth. Given that we fully inline virt_to_page and friends now that is rather good. total used free shared buffers cached Mem: 8264720 57240 8207480 0 352 11516 -/+ buffers/cache: 45372 8219348 Swap: 9775512 0 9775512 The total available memory is increased by 8k. This patch makes sparsemem the default and removes discontig and flatmem support from x86. [ akpm@linux-foundation.org: allnoconfig build fix ] Acked-by:Andi Kleen <ak@suse.de> Signed-off-by:
Christoph Lameter <clameter@sgi.com> Signed-off-by:
-
Borislav Petkov authored
Remove repeated comment from the linker script for the x86-32 target. Signed-off-by:
Borislav Petkov <bbpetkov@yahoo.de> Signed-off-by:
-
Yinghai Lu authored
In init/main.c boot_cpu_init() does that later. Signed-off-by:
Yinghai Lu <yinghai.lu@sun.com> Cc: Zachary Amsden <zach@vmware.com> Cc: Andi Kleen <ak@suse.de> Signed-off-by:
-
Yinghai Lu authored
Some BIOSes that support two/four dualcore/quadcore systems, will get: ACPI: LAPIC (acpi_id[0x01] lapic_id[0x00] enabled) Processor #0 15:1 APIC version 16 ACPI: LAPIC (acpi_id[0x02] lapic_id[0x01] enabled) Processor #1 15:1 APIC version 16 ACPI: LAPIC (acpi_id[0x03] lapic_id[0x02] enabled) Processor #2 15:1 APIC version 16 ACPI: LAPIC (acpi_id[0x04] lapic_id[0x03] enabled) Processor #3 15:1 APIC version 16 ACPI: LAPIC (acpi_id[0x05] lapic_id[0x84] disabled) ACPI: LAPIC (acpi_id[0x06] lapic_id[0x85] disabled) ACPI: LAPIC (acpi_id[0x07] lapic_id[0x86] disabled) ACPI: LAPIC (acpi_id[0x08] lapic_id[0x87] disabled) ACPI: LAPIC (acpi_id[0x09] lapic_id[0x88] disabled) ACPI: LAPIC (acpi_id[0x0a] lapic_id[0x89] disabled) ACPI: LAPIC (acpi_id[0x0b] lapic_id[0x8a] disabled) ACPI: LAPIC (acpi_id[0x0c] lapic_id[0x8b] disabled) ACPI: LAPIC (acpi_id[0x0d] lapic_id[0x8c] disabled) ACPI: LAPIC (acpi_id[0x0e] lapic_id[0x8d] disabled) ACPI: LAPIC (acpi_id[0x0f] lapic_id[0x8e] disabled) ACPI: LAPIC (acpi_id[0x10] lapic_id[0x8f] disabled) SMP: Allowing 16 CPUs, 12 hotplug CPUs the /proc/cpuinfo will show a bunch of NULL cpus with cpu_index=0 so assign impossible cpu_index value at first instead of 0. Signed-off-by:
-
Vladimir Berezniker authored
Sanitize user specified e820 memory ranges, using the same logic that is applied to the values returned by the BIOS. This ensures consistent handling regardless of the source of the memory mappings. Allows overriding portions of the memory map without specifying one in it's entirety (memmap=exactmap). E.g. marking a range of bad RAM as reserved with memmap=48M$528M BIOS supplied range BIOS-e820: 0000000000100000 - 000000007fe80000 (usable) becomes user: 0000000000100000 - 0000000021000000 (usable) user: 0000000021000000 - 0000000024000000 (reserved) user: 0000000024000000 - 000000007fe80000 (usable) Previously this did not work, as the original BIOS range was left untouched while the user defined range was appended to the end of the memory map. [ tglx: arch/x86 adaptation ] Signed-off-by:
Vladimir Berezniker <vmpn@hitechman.com> Signed-off-by:
-
Roland McGrath authored
This consolidates the four different places that implemented the same encoding magic for the GDT-slot 32-bit TLS support. The old tls32.c was renamed and is now only slightly modified to be the shared implementation. Signed-off-by:
Roland McGrath <roland@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Zachary Amsden <zach@vmware.com> Signed-off-by:
-
Roland McGrath authored
This renames arch/x86/ia32/tls32.c to arch/x86/kernel/tls.c, which does nothing now but paves the way to consolidate this code for 32-bit too. Signed-off-by:
-
Roland McGrath authored
This replaces the desc_empty macro with an inline. It now handles easily any of the four different types used between 32/64 code to refer to these 8 bytes. It's identical in both asm-x86/processor_64.h and asm-x86/processor_32.h, so if these files ever get merged this function can be in the common code. This also removes the desc_equal macro because nothing uses it. Signed-off-by:
-
Roland McGrath authored
The fs_base and gs_base fields are available in user_regs_struct. But reading these via ptrace (PTRACE_GETREGS or PTRACE_PEEKUSR) does not give a reliably useful value. The thread_struct fields are 0 when do_arch_prctl decided to use a GDT slot instead of MSR_FS_BASE, which it does for a value under 1<<32. This changes ptrace access to fs_base and gs_base to work like PTRACE_ARCH_PRCTL does. That is, it reads the base address that user-mode memory access using the fs/gs instruction prefixes will use, regardless of how it's being implemented in the kernel. The MSR vs GDT is an implementation detail that is pretty much hidden from userland in the actual using, and there is no reason that ptrace should give the internal implementation picture rather than the user-mode semantic picture. In the case of setting the value, this can implicitly change the fsindex/gsindex value (also separately in user_regs_struct), which is what happens when the thread calls arch_prctl itself. In a PTRACE_SETREGS, the fs_base change will come after the fsindex change due to the order of the struct, and so a change the debugger made to fs_base will have the effect intended, another part of the user_regs_struct will now differ when read back from what the debugger wrote. This makes PTRACE_ARCH_PRCTL obsolete. We could consider declaring it deprecated and removing it one day, though there is no hurry. For the foreseeable future, debuggers have to assume an old kernel that does not report reliable fs_base/gs_base values in user_regs_struct and stick to PTRACE_ARCH_PRCTL anyway. Signed-off-by:
-
Roland McGrath authored
This changes a couple of places to use the get_desc_base function. They were duplicating the same calculation with different equivalent code. Signed-off-by:
-
Roland McGrath authored
This defines the get_desc_base function in asm-x86/desc_64.h to match the one in desc_32.h. If these two files ever get merged together, this function could be the same in both. Signed-off-by:
-
Roland McGrath authored
Some assembler versions automagically optimize .eh_frame contents, changing their size. The CFI in sysenter.S was not using optimal formatting, so it would be changed by newer/smarter assemblers. This ran afoul of the wired constant for padding out the other vDSO images to match its size. This changes the original hand-coded source to use the optimal format encoding for its operations. That leaves nothing more for a fancy assembler to do, so the sizes will match the wired-in expected size regardless of the assembler version. Signed-off-by:
-
Roland McGrath authored
This cleans up the arch/x86/vdso/Makefile rules for vdso.so to share more code with the vdso32-*.so rules and remove old cruft. Signed-off-by:
-
Roland McGrath authored
This reorders the code in the 32-bit vDSO images to put the signal trampolines first and __kernel_vsyscall after them. The order does not matter to userland, it just uses what AT_SYSINFO or e_entry says. Since the signal trampolines are the same size in both versions of the vDSO, putting them first is the simplest way to get the addresses to line up. This makes it work to use a more compact layout for the vDSO. Signed-off-by:
-
Roland McGrath authored
This removes all the old vsyscall code from arch/x86/ia32/ that is no longer used because arch/x86/vdso/ code has replaced it. Signed-off-by:
-
Roland McGrath authored
This makes x86_64's ia32 emulation support share the sources used in the 32-bit kernel for the 32-bit vDSO and much of its setup code. The 32-bit vDSO mapping now behaves the same on x86_64 as on native 32-bit. The abi.syscall32 sysctl on x86_64 now takes the same values that vm.vdso_enabled takes on the 32-bit kernel. That is, 1 means a randomized vDSO location, 2 means the fixed old address. The CONFIG_COMPAT_VDSO option is now available to make this the default setting, the same meaning it has for the 32-bit kernel. (This does not affect the 64-bit vDSO.) The argument vdso32=[012] can be used on both 32-bit and 64-bit kernels to set this paramter at boot time. The vdso=[012] argument still does this same thing on the 32-bit kernel. Signed-off-by:
-
Roland McGrath authored
This puts the syscall version of the 32-bit vDSO in arch/x86/vdso/vdso32/ for 64-bit IA32 support. This is not used yet, but it paves the way for consolidating the 32-bit vDSO source and build logic all in one place. Signed-off-by:
-
Roland McGrath authored
This changes the 64-bit kernel's support for the 32-bit sysenter instruction to use stored fields rather than constants for the user-mode return address, as the 32-bit kernel does. This adds a sysenter_return field to struct thread_info, as 32-bit has. There is no observable effect from this yet. It makes the assembly code independent of the 32-bit vDSO mapping address, paving the way for making the vDSO address vary as it does on the 32-bit kernel. [ akpm@linux-foundation.org: build fix on !CONFIG_IA32_EMULATION ] Signed-off-by:
-
Roland McGrath authored
This harmonizes the name for the entry point from the 32-bit sysenter instruction across 32-bit and 64-bit kernels. Signed-off-by:
-
Roland McGrath authored
This moves arch/x86/kernel/sysenter_32.c to arch/x86/vdso/vdso32-setup.c, keeping all the code relating only to vDSO magic in the vdso/ subdirectory. This is a pure renaming, but it paves the way to consolidating the code for dealing with 32-bit vDSOs across CONFIG_X86_32 and CONFIG_IA32_EMULATION. Signed-off-by:
-
Roland McGrath authored
This enables 'make vdso_install' for i386 as on x86_64 and powerpc. Signed-off-by:
-
Roland McGrath authored
This updates the exceptions for absolute relocs for the new symbol name convention used for symbols extracted from the vDSO images. Signed-off-by:
-
Roland McGrath authored
This makes the i386 kernel use the new vDSO build in arch/x86/vdso/vdso32/ to replace the old one from arch/x86/kernel/. Signed-off-by:
-
Roland McGrath authored
This builds the 32-bit vDSO images in the arch/x86/vdso subdirectory. Nothing uses the images yet, but this paves the way for consolidating the vDSO build logic all in one place. The new images use a linker script sharing the layout parts from vdso-layout.lds.S with the 64-bit vDSO. A new vdso32-syms.lds is generated in the style of vdso-syms.lds. Signed-off-by:
-
Roland McGrath authored
This moves the i386 vDSO sources into arch/x86/vdso/vdso32/, a new directory. This patch is a pure renaming, but paves the way for consolidating the vDSO build logic. Signed-off-by:
-
Roland McGrath authored
This change harmonizes the asm-offsets macros used in the 32-bit vDSO across 32-bit and 64-bit builds. It's a purely cosmetic change for now, but it paves the way for consolidating the 32-bit vDSO builds. Signed-off-by:
-
Roland McGrath authored
This revamps the vDSO linker script to lay things out with the best packing of the data and good, separate alignment of the code. The rigid layout using VDSO_TEXT_OFFSET no longer matters to the kernel. I've moved the layout parts of the linker script into a new include file, vdso-layout.lds.S; this is in preparation for sharing the script for the 32-bit vDSO builds too. Signed-off-by:
-
Roland McGrath authored
Get rid of vdso-syms.o from the kernel link. We don't need it any more. Signed-off-by:
-
Roland McGrath authored
This patch changes the kernel's references to addresses in the vDSO image to be based on the symbols defined by vdso-syms.lds instead of the old vdso-syms.o symbols. This is all wrapped up in a macro defined by the new asm-x86/vdso.h header; that's the only place in the kernel source that has to know the details of the scheme for getting vDSO symbol values. Signed-off-by:
-
Roland McGrath authored
This patch adds a new way of extracting symbols from the built vDSO image. This is much simpler and less fragile than using ld -R; it removes the need to control the DSO layout quite so exactly. I was clearly unduly distracted by clever ld uses when I did the original vDSO implementation. Signed-off-by:
-
Jiri Kosina authored
Randomize the location of the heap (brk) for i386 and x86_64. The range is randomized in the range starting at current brk location up to 0x02000000 offset for both architectures. This, together with pie-executable-randomization.patch and pie-executable-randomization-fix.patch, should make the address space randomization on i386 and x86_64 complete. Arjan says: This is known to break older versions of some emacs variants, whose dumper code assumed that the last variable declared in the program is equal to the start of the dynamically allocated memory region. (The dumper is the code where emacs effectively dumps core at the end of it's compilation stage; this coredump is then loaded as the main program during normal use) iirc this was 5 years or so; we found this way back when I was at RH and we first did the security stuff there (including this brk randomization). It wasn't all variants of emacs, and it got fixed as a result (I vaguely remember that emacs already had code to deal with it for other archs/oses, just ifdeffed wrongly). It's a rare and wrong assumption as a general thing, just on x86 it mostly happened to be true (but to be honest, it'll break too if gcc does something fancy or if the linker does a non-standard order). Still its something we should at least document. Note 2: afaik it only broke the emacs *build*. I'm not 100% sure about that (it IS 5 years ago) though. [ akpm@linux-foundation.org: deuglification ] Signed-off-by:
Jiri Kosina <jkosina@suse.cz> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Jakub Jelinek <jakub@redhat.com> Signed-off-by:
-
Robert Richter authored
Also macro definitions in apicdef.h has been updated. Signed-off-by:
-
Andi Kleen authored
Earlier patch added IO APIC setup into local APIC setup. This caused modpost warnings. Fix them by untangling setup_local_APIC() and splitting it into smaller functions. The IO APIC initialization is only called for the BP init. Also removed some outdated debugging code and minor cleanup. [ tglx: arch/x86 adaptation ] Signed-off-by:
-
Yinghai Lu authored
We shoud use core id bits instead of max cores, in case later with AMD downcores Quad core Opteron. [ tglx: arch/x86 adaptation ] Signed-off-by:
Yinghai Lu <yhlu.kernel@gmail.com> Signed-off-by:
-
Yinghai Lu authored
We need to store core id bits to cpuinfo_x86 in early_identify_cpu. So we use it to create acpiid_to_node array in k8topolgy.c Signed-off-by:
-
Adrian Bunk authored
Contrary to the comment "newer gccs do it by default", newer gcc versions default to -maccumulate-outgoing-args only with CONFIG_CC_OPTIMIZE_FOR_SIZE=n, and then only with some CPU settings. Measured with an i386 defconfig, gcc 4.2.1 and kernel 2.6.23-rc1 ("orig" is the plain kernel, "changed is with -maccumulate-outgoing-args removed): $ ls -la vmlinux* -rwxrwxr-x 1 bunk bunk 6269713 2007-07-24 22:19 vmlinux.changed -rwxrwxr-x 1 bunk bunk 6425361 2007-07-24 22:19 vmlinux.orig $ size vmlinux.* text data bss dec hex filename 4493465 504108 614400 5611973 55a1c5 vmlinux.changed 4646160 504108 614400 5764668 57f63c vmlinux.orig $ That's a 2.5% size increase that does for sure hurt small systems. If the stack unwinder ever comes back and needs this as indicated in the comment, adding it to the cflags when the user enabled the unwinder should be a better option. [ tglx: arch/x86 adaptation ] Signed-off-by:Adrian Bunk <bunk@stusta.de> Signed-off-by:
-
Yinghai Lu authored
4 socket quad core, 8 socket quad core will do apic ID lifting for BSP. But io-apic regs for ExtINT still use 0 as dest. so when we enable apic error vector in BSP, we will get one APIC error. CPU: L1 I Cache: 64K (64 bytes/line), D cache 64K (64 bytes/line) CPU: L2 Cache: 512K (64 bytes/line) CPU 0/4 -> Node 0 CPU: Physical Processor ID: 1 CPU: Processor Core ID: 0 SMP alternatives: switching to UP code ACPI: Core revision 20070126 enabled ExtINT on CPU#0 ESR value after enabling vector: 00000000, after 0000000c APIC error on CPU0: 0c(08) ENABLING IO-APIC IRQs Synchronizing Arb IDs. So move enable_IO_APIC from setup_IO_APIC into setup_local_APIC and call it before enabling the ACPI error vector. [ tglx: arch/x86 adaptation ] Signed-off-by:
-
Thomas Gleixner authored
Clean it up before applying more patches to it. Signed-off-by:
-
