| Message ID | 20210202002109.GA624440@juliacomputing.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | arm64: ptrace: Fix missing return in hw breakpoint code | expand |
Hi Keno, On Mon, Feb 01, 2021 at 07:21:09PM -0500, Keno Fischer wrote: > This was found by inspection - I don't work with aarch32 very > much, but I believe there is a missing return in this code, > since the second force_sig_fault here would override the > first, thus making the whole compat code here pointless. > > Signed-off-by: Keno Fischer <keno@juliacomputing.com> > --- > > This was left over in my working tree from when I was working > through various aarch64 ptrace weirdness last year and I don't > think I ever got around to submitting it, but I just looked > through my work from back then and came across this again - > my apologies if I'm wrong about this - it's been a while, > so I'm not really read into this code anymore. > > arch/arm64/kernel/ptrace.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/arch/arm64/kernel/ptrace.c b/arch/arm64/kernel/ptrace.c > index e7b01904f180..9309c41dbafe 100644 > --- a/arch/arm64/kernel/ptrace.c > +++ b/arch/arm64/kernel/ptrace.c > @@ -195,6 +195,7 @@ static void ptrace_hbptriggered(struct perf_event *bp, > arm64_force_sig_ptrace_errno_trap(si_errno, > (void __user *)bkpt->trigger, > desc); > + return; > } > #endif > arm64_force_sig_fault(SIGTRAP, TRAP_HWBKPT, Thanks, I'll apply this for 5.12. However, I was trying to rework the commit message to describe the impact of the issue and I'm not sure I agree with your assessment that the second signal overrides the first. My (admittedly brief) reading of __send_signal() is that the subsequent signal is effectively ignored. Am I missing something? Cheers, Will
Hi Will, > Thanks, I'll apply this for 5.12. However, I was trying to rework the commit > message to describe the impact of the issue and I'm not sure I agree with > your assessment that the second signal overrides the first. My (admittedly > brief) reading of __send_signal() is that the subsequent signal is > effectively ignored. Am I missing something? Ah, I suppose you're probably right. I hadn't looked at this code in quite a while, so I must have mis-remembered what the impact was. Keno
diff --git a/arch/arm64/kernel/ptrace.c b/arch/arm64/kernel/ptrace.c index e7b01904f180..9309c41dbafe 100644 --- a/arch/arm64/kernel/ptrace.c +++ b/arch/arm64/kernel/ptrace.c @@ -195,6 +195,7 @@ static void ptrace_hbptriggered(struct perf_event *bp, arm64_force_sig_ptrace_errno_trap(si_errno, (void __user *)bkpt->trigger, desc); + return; } #endif arm64_force_sig_fault(SIGTRAP, TRAP_HWBKPT,
This was found by inspection - I don't work with aarch32 very much, but I believe there is a missing return in this code, since the second force_sig_fault here would override the first, thus making the whole compat code here pointless. Signed-off-by: Keno Fischer <keno@juliacomputing.com> --- This was left over in my working tree from when I was working through various aarch64 ptrace weirdness last year and I don't think I ever got around to submitting it, but I just looked through my work from back then and came across this again - my apologies if I'm wrong about this - it's been a while, so I'm not really read into this code anymore. arch/arm64/kernel/ptrace.c | 1 + 1 file changed, 1 insertion(+)