Re: [PATCH] elf, coredump: Extract only the active register set during core dump

[Pedro Alves]

On 05/23/2014 06:16 AM, Anshuman Khandual wrote:
> Regset active hooks provide a way to query how many registers in the
> register set are active at any point of time. Currently this information
> is being ignored while creating core dump sections corresponding to any
> core note register set. This way the core dump will contain data which are
> not part of the active context of the process and may not be useful. This
> patch will make sure that only the active part of the register set are
> captured during the core dump process which will reduce the core dump
> size.
> 
> Signed-off-by: Anshuman Khandual <khandual@xxxxxxxxxxxxxxxxxx>
> ---
> NOTE:
> Pedro Alves has mentioned that producing smaller note sections in the core
> dump may break some existing consumers. I request suggestions, reviews and
> test reports on different architectures to prove that this patch does not
> break any existing consumer. Thank you.

Yeah, FYI, I mentioned that after noticing that ia64 does:

mainline/linux-2.6/arch/ia64/kernel/ptrace.c:

 static int
 fpregs_active(struct task_struct *target, const struct user_regset *regset)
 {
         return (target->thread.flags & IA64_THREAD_FPH_VALID) ? 128 : 32;
 }

And it's likely that tools expect fpregset_t to have a fixed size.

include/uapi/linux/elfcore.h:

 22 typedef elf_fpregset_t fpregset_t;

arch/ia64/include/asm/elf.h:

 186 typedef struct ia64_fpreg elf_fpreg_t;
 187 typedef elf_fpreg_t elf_fpregset_t[ELF_NFPREG];

arch/ia64/include/asm/elf.h:

 154 #define ELF_NGREG       128     /* we really need just 72 but let's leave some headroom... */
 155 #define ELF_NFPREG      128     /* f0 and f1 could be omitted, but so what... */


I haven't done an exhaustive look over ports.

-- 
Pedro Alves

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/