mbox series

[for-9.1,v2,00/28] linux-user/i386: Properly align signal frame

Message ID 20240409050302.1523277-1-richard.henderson@linaro.org
Headers show
Series linux-user/i386: Properly align signal frame | expand

Message

Richard Henderson April 9, 2024, 5:02 a.m. UTC 
v1: https://lore.kernel.org/qemu-devel/20230524054647.1093758-1-richard.henderson@linaro.org/

But v1 isn't particularly complet or korrect.

Disconnect fpstate from sigframe, just like the kernel does.
Return the separate portions of the frame from get_sigframe.
Alter all of the target fpu routines to access memory that
has already been translated and sized.


r~


Richard Henderson (28):
  target/i386: Add tcg/access.[ch]
  target/i386: Convert do_fldt, do_fstt to X86Access
  target/i386: Convert helper_{fbld,fbst}_ST0 to X86Access
  target/i386: Convert do_fldenv to X86Access
  target/i386: Convert do_fstenv to X86Access
  target/i386: Convert do_fsave, do_frstor to X86Access
  target/i386: Convert do_xsave_{fpu,mxcr,sse} to X86Access
  target/i386: Convert do_xrstor_{fpu,mxcr,sse} to X86Access
  tagret/i386: Convert do_fxsave, do_fxrstor to X86Access
  target/i386: Convert do_xsave_* to X86Access
  target/i386: Convert do_xrstor_* to X86Access
  target/i386: Split out do_xsave_chk
  target/i386: Add rbfm argument to cpu_x86_{xsave,xrstor}
  target/i386: Add {hw,sw}_reserved to X86LegacyXSaveArea
  linux-user/i386: Drop xfeatures_size from sigcontext arithmetic
  linux-user/i386: Remove xfeatures from target_fpstate_fxsave
  linux-user/i386: Replace target_fpstate_fxsave with X86LegacyXSaveArea
  linux-user/i386: Split out struct target_fregs_state
  linux-user/i386: Fix -mregparm=3 for signal delivery
  linux-user/i386: Return boolean success from restore_sigcontext
  linux-user/i386: Return boolean success from xrstor_sigcontext
  linux-user/i386: Fix allocation and alignment of fp state
  target/i386: Honor xfeatures in xrstor_sigcontext
  target/i386: Convert do_xsave to X86Access
  target/i386: Convert do_xrstor to X86Access
  target/i386: Pass host pointer and size to cpu_x86_{fsave,frstor}
  target/i386: Pass host pointer and size to cpu_x86_{fxsave,fxrstor}
  target/i386: Pass host pointer and size to cpu_x86_{xsave,xrstor}

 target/i386/cpu.h                |  57 ++-
 target/i386/tcg/access.h         |  40 ++
 linux-user/i386/signal.c         | 669 ++++++++++++++++++-------------
 target/i386/tcg/access.c         | 160 ++++++++
 target/i386/tcg/fpu_helper.c     | 561 ++++++++++++++++----------
 tests/tcg/x86_64/test-1648.c     |  33 ++
 target/i386/tcg/meson.build      |   1 +
 tests/tcg/x86_64/Makefile.target |   1 +
 8 files changed, 1014 insertions(+), 508 deletions(-)
 create mode 100644 target/i386/tcg/access.h
 create mode 100644 target/i386/tcg/access.c
 create mode 100644 tests/tcg/x86_64/test-1648.c

Comments

Paolo Bonzini April 9, 2024, 7:52 a.m. UTC | #1 
On 4/9/24 07:02, Richard Henderson wrote:
> v1: https://lore.kernel.org/qemu-devel/20230524054647.1093758-1-richard.henderson@linaro.org/
> 
> But v1 isn't particularly complet or korrect.
> 
> Disconnect fpstate from sigframe, just like the kernel does.
> Return the separate portions of the frame from get_sigframe.
> Alter all of the target fpu routines to access memory that
> has already been translated and sized.

With the exception of patch 22, and with small nits in patches 1/19/23:

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>

> 
> r~
> 
> 
> Richard Henderson (28):
>    target/i386: Add tcg/access.[ch]
>    target/i386: Convert do_fldt, do_fstt to X86Access
>    target/i386: Convert helper_{fbld,fbst}_ST0 to X86Access
>    target/i386: Convert do_fldenv to X86Access
>    target/i386: Convert do_fstenv to X86Access
>    target/i386: Convert do_fsave, do_frstor to X86Access
>    target/i386: Convert do_xsave_{fpu,mxcr,sse} to X86Access
>    target/i386: Convert do_xrstor_{fpu,mxcr,sse} to X86Access
>    tagret/i386: Convert do_fxsave, do_fxrstor to X86Access
>    target/i386: Convert do_xsave_* to X86Access
>    target/i386: Convert do_xrstor_* to X86Access
>    target/i386: Split out do_xsave_chk
>    target/i386: Add rbfm argument to cpu_x86_{xsave,xrstor}
>    target/i386: Add {hw,sw}_reserved to X86LegacyXSaveArea
>    linux-user/i386: Drop xfeatures_size from sigcontext arithmetic
>    linux-user/i386: Remove xfeatures from target_fpstate_fxsave
>    linux-user/i386: Replace target_fpstate_fxsave with X86LegacyXSaveArea
>    linux-user/i386: Split out struct target_fregs_state
>    linux-user/i386: Fix -mregparm=3 for signal delivery
>    linux-user/i386: Return boolean success from restore_sigcontext
>    linux-user/i386: Return boolean success from xrstor_sigcontext
>    linux-user/i386: Fix allocation and alignment of fp state
>    target/i386: Honor xfeatures in xrstor_sigcontext
>    target/i386: Convert do_xsave to X86Access
>    target/i386: Convert do_xrstor to X86Access
>    target/i386: Pass host pointer and size to cpu_x86_{fsave,frstor}
>    target/i386: Pass host pointer and size to cpu_x86_{fxsave,fxrstor}
>    target/i386: Pass host pointer and size to cpu_x86_{xsave,xrstor}
> 
>   target/i386/cpu.h                |  57 ++-
>   target/i386/tcg/access.h         |  40 ++
>   linux-user/i386/signal.c         | 669 ++++++++++++++++++-------------
>   target/i386/tcg/access.c         | 160 ++++++++
>   target/i386/tcg/fpu_helper.c     | 561 ++++++++++++++++----------
>   tests/tcg/x86_64/test-1648.c     |  33 ++
>   target/i386/tcg/meson.build      |   1 +
>   tests/tcg/x86_64/Makefile.target |   1 +
>   8 files changed, 1014 insertions(+), 508 deletions(-)
>   create mode 100644 target/i386/tcg/access.h
>   create mode 100644 target/i386/tcg/access.c
>   create mode 100644 tests/tcg/x86_64/test-1648.c
>