This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [Patch] MIPS: Fix unwinding through signal frames.


David Daney <ddaney@avtrex.com> writes:
> Certain MIPS processors have an errata workaround that changes the stack
> layout for signal handlers.  The current MD_FALLBACK_FRAME_STATE_FOR for
> mips/Linux assumes that the signal context is at a constant offset from
> the signal trampoline.  When the workaround is in effect, the signal
> context is not in the expected location, so unwinding fails.
>
> I discovered this problem while running the testsuite on an SGI O2
> system.  All the other mips systems I test with don't have the
> workaround so were not effected by the bug.
>
> The patch changes the calculation of the signal context so that it uses
> the CFA of the unwind context to find the signal context.  This has the
> advantage of working on system either with or without the workaround.
>
> Tested on mipsel-linux (without workaround) and spot checked on
> mips64-linux n32 ABI (with workaround) with full mips64-linux test in
> progress.
>
> OK to commit?

Yup.  Thanks for working on this.

> +	u_int32_t ass[4];

Can you put "/* Argument save space for o32.  */" next to this,
as I fear the name isn't self-descriptive...

Richard


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]