lkml.org 
[lkml]   [2015]   [Feb]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH 01/11] x86: entry_64.S: always allocate complete "struct pt_regs"
On Wed, Jan 14, 2015 at 1:48 PM, Denys Vlasenko <dvlasenk@redhat.com> wrote:
> 64-bit code was using six stack slots less by not saving/restoring
> registers which are callee-preserved according to C ABI,
> and not allocating space for them.
> Only when syscall needed a complete "struct pt_regs",
> the complete area was allocated and filled in.
> As an additional twist, on interrupt entry a "slightly less truncated pt_regs"
> trick is used, to make nested interrupt stacks easier to unwind.
>
> This proved to be a source of significant obfuscation and subtle bugs.
> For example, stub_fork had to pop the return address,
> extend the struct, save registers, and push return address back. Ugly.
> ia32_ptregs_common pops return address and "returns" via jmp insn,
> throwing a wrench into CPU return stack cache.
>
> This patch changes code to always allocate a complete "struct pt_regs".
> The saving of registers is still done lazily.
>
> "Partial pt_regs" trick on interrupt stack is retained, but with added comments
> which explain what we are doing, and why. Existing comments are improved.
>
> Macros which manipulate "struct pt_regs" on stack are reworked:
> ALLOC_PT_GPREGS_ON_STACK allocates the structure.
> SAVE_C_REGS saves to it those registers which are clobbered by C code.
> SAVE_EXTRA_REGS saves to it all other registers.
> Corresponding RESTORE_* and REMOVE_PT_GPREGS_FROM_STACK macros reverse it.
>
> ia32_ptregs_common, stub_fork and friends lost their ugly dance with
> return pointer.
>
> LOAD_ARGS32 in ia32entry.S now uses symbolic stack offsets
> instead of magic numbers.
>
> error_entry and save_paranoid now use SAVE_C_REGS + SAVE_EXTRA_REGS
> instead of having it open-coded yet again.
>
> Misleading and slightly wrong comments in "struct pt_regs" are fixed
> (four instances).
>
> Patch was run-tested: 64-bit executables, 32-bit executables,
> strace works.
> Timing tests did not show measurable difference in 32-bit
> and 64-bit syscalls.

I like this. However, can you split it? I think it would be easier
to review and more bisect-friendly if there were first a patch to do
all the macro and comment cleanup without any layout changes followed
by bite-sized layout changes. As it stands, I can't just go through
this to make sure that nothing is changing, because things are
changing.

--Andy


\
 
 \ /
  Last update: 2015-02-11 21:41    [W:0.624 / U:0.096 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site