Messages in this thread | | | From | Andy Lutomirski <> | Date | Thu, 23 Jun 2016 14:53:18 -0700 | Subject | Re: [PATCH v3 3/3] x86/ptrace, x86/signal: Remove TS_I386_REGS_POKED |
| |
On Thu, Jun 23, 2016 at 2:26 PM, Oleg Nesterov <oleg@redhat.com> wrote: > Again, I think the patch is fine, just a question > > On 06/20, Andy Lutomirski wrote: >> >> System call restart has some oddities wrt ptrace: >> >> 1. For whatever reason, the kernel delivers signals and triggers >> ptrace before handling syscall restart. This means that >> -ERESTART_RESTARTBLOCK, etc is visible to userspace. We could >> plausibly get away with changing that, but it seems quite risky. > > How we can change this? > > The kernel simply can't know how it should react to (say) -ERESTARTSYS > until debugger acks/nacks/changes the signal reported by tracee.
Hmm, good point. I don't know whether our current behavior is fully correct or whether we could change it.
> >> + /* >> + * A 32-bit ptracer has the following expectations: >> + * >> + * - Storing -1 (i.e. 0xffffffff) to orig_eax will prevent >> + * syscall restart handling. >> + * >> + * - Restoring regs saved on exit from an interrupted >> + * restartable syscall will trigger syscall restart. Such >> + * regs will have non-negative orig_eax and negative eax. >> + * >> + * The kernel's syscall restart code treats regs->orig_ax and >> + * regs->ax as 64-bit signed quantities. 32-bit user code >> + * doesn't care about the high bits. Keep it simple and just >> + * sign-extend both values. >> + */ >> + R32_SIGNED(orig_eax, orig_ax); >> + R32_SIGNED(eax, ax); > > OK. but do we really need R32_SIGNED(orig_eax) ? syscall_get_nr() > returns "int", not "long".
Fair enough, maybe we don't. I'll drop that part and just keep R32_SIGNED(eax, ax).
--Andy
| |