[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH v5] tcg-aarch64: Properly detect SIGSEGV writes
From: |
Claudio Fontana |
Subject: |
Re: [Qemu-devel] [PATCH v5] tcg-aarch64: Properly detect SIGSEGV writes |
Date: |
Wed, 16 Apr 2014 17:40:09 +0200 |
User-agent: |
Mozilla/5.0 (Windows NT 6.1; rv:24.0) Gecko/20100101 Thunderbird/24.0.1 |
On 16.04.2014 17:14, Richard Henderson wrote:
> Since the kernel doesn't pass any info on the reason for the fault,
> disassemble the instruction to detect a store.
>
> Signed-off-by: Richard Henderson <address@hidden>
> ---
>
> Rather than spam the rest of the reviewed patch series to the list,
> I'm just sending the update of the one patch that had issues.
>
> From v4, Claudio correctly diagnosed that I had missed the 128-bit
> stores. Unfortunately, opcode 2 is used for both loads and stores,
> so one has to use a different mask for them.
>
> Previous revisions had a comment about how some of these tests could
> be merged. Adding another 7 for the 128-bit stores seemed a bridge
> too far, so I did some easy merging.
>
> There are two groups of insns that differ only in pre-/post-/no-increment,
> or that vary only in fields that don't matter (register input vs immediate
> input). I no longer try to distinguish those.
>
> In the case of C3.3.8-12, there are combinations of bits 10, 11, 21 that
> are not valid insns. But since we don't really need to be a full
> disassembler I think that'll be fine for now.
>
> With any luck the kernel patch that passes the relevent data down from
> the trap will be accepted upstream soon, and this will just be legacy.
>
>
> r~
>
> ---
> user-exec.c | 25 +++++++++++++++++++------
> 1 file changed, 19 insertions(+), 6 deletions(-)
>
> diff --git a/user-exec.c b/user-exec.c
> index bc58056..8ed6fec 100644
> --- a/user-exec.c
> +++ b/user-exec.c
> @@ -465,16 +465,29 @@ int cpu_signal_handler(int host_signum, void *pinfo,
>
> #elif defined(__aarch64__)
>
> -int cpu_signal_handler(int host_signum, void *pinfo,
> - void *puc)
> +int cpu_signal_handler(int host_signum, void *pinfo, void *puc)
> {
> siginfo_t *info = pinfo;
> struct ucontext *uc = puc;
> - uint64_t pc;
> - int is_write = 0; /* XXX how to determine? */
> + uintptr_t pc = uc->uc_mcontext.pc;
> + uint32_t insn = *(uint32_t *)pc;
> + bool is_write;
>
> - pc = uc->uc_mcontext.pc;
> - return handle_cpu_signal(pc, (uint64_t)info->si_addr,
> + /* XXX: need kernel patch to get write flag faster. */
> + is_write = ( (insn & 0xbfff0000) == 0x0c000000 /* C3.3.1 */
> + || (insn & 0xbfe00000) == 0x0c800000 /* C3.3.2 */
> + || (insn & 0xbfdf0000) == 0x0d000000 /* C3.3.3 */
> + || (insn & 0xbfc00000) == 0x0d800000 /* C3.3.4 */
> + || (insn & 0x3f400000) == 0x08000000 /* C3.3.6 */
> + || (insn & 0x3bc00000) == 0x39000000 /* C3.3.13 */
> + || (insn & 0x3fc00000) == 0x3d800000 /* ... 128bit */
> + /* Ingore bits 10, 11 & 21, controlling indexing. */
> + || (insn & 0x3bc00000) == 0x38000000 /* C3.3.8-12 */
> + || (insn & 0x3fe00000) == 0x3c800000 /* ... 128bit */
> + /* Ignore bits 23 & 24, controlling indexing. */
> + || (insn & 0x3a400000) == 0x28000000); /* C3.3.7,14-16 */
> +
> + return handle_cpu_signal(pc, (uintptr_t)info->si_addr,
> is_write, &uc->uc_sigmask, puc);
> }
>
>
Reviewed-by: Claudio Fontana <address@hidden>