. Consequently, seccomp_bpf selftest passes both with and
> > without this change.
> >
> > Now comes the moment when PTRACE_SET_SYSCALL_INFO is going to be
> > introduced. PTRACE_SET_SYSCALL_INFO is a generic ptrace API that
> > complements PTRACE_GET_SYSCALL_INFO by letting the ptracer modify
> > the details of the system calls the tracee is blocked in.
> >
> > One of the helpers that have to be used to implement
> > PTRACE_SET_SYSCALL_INFO is syscall_set_return_value().
> > This helper complements other two helpers, syscall_get_error() and
> > syscall_get_return_value(), that are currently used to implement
> > PTRACE_GET_SYSCALL_INFO on syscall return. When syscall_set_return_value()
> > is used to set an error code, the caller specifies it as a negative value
> > in -ERRORCODE format.
> >
> > Unfortunately, this does not work well on powerpc since commit 1b1a3702a65c
> > ("powerpc: Don't negate error in syscall_set_return_value()") because
> > syscall_set_return_value() does not follow the powerpc sc syscall return
> > ABI:
> > /*
> >* In the general case it's not obvious that we must deal with
> >* CCR here, as the syscall exit path will also do that for us.
> >* However there are some places, eg. the signal code, which
> >* check ccr to decide if the value in r3 is actually an error.
> >*/
> > if (error) {
> > regs->ccr |= 0x1000L;
> > regs->gpr[3] = error;
> > } else {
> > regs->ccr &= ~0x1000L;
> > regs->gpr[3] = val;
> > }
> >
> > The reason why this syscall_set_return_value() implementation was able to
> > get away with violating the powerpc sc syscall return ABI is the following:
> > Up to now, syscall_set_return_value() on powerpc could be called only from
> > do_syscall_trace_enter() via do_seccomp(), there was no way it could be
> > called from do_syscall_trace_leave() which is the point where tracers on
> > syscall return are activated and the powerpc sc syscall return ABI has
> > to be respected.
> >
> > Introduction of PTRACE_SET_SYSCALL_INFO necessitates a change of
> > syscall_set_return_value() to comply with the powerpc sc syscall return
> > ABI. Without the change, the upcoming ptrace/set_syscall_info selftest
> > fails with the following diagnostics:
> >
> ># set_syscall_info.c:119:set_syscall_info:Expected exp_exit->rval (-38)
> > == info->exit.rval (38)
> ># set_syscall_info.c:120:set_syscall_info:wait #4:
> > PTRACE_GET_SYSCALL_INFO #2: exit stop mismatch
> >
> > Note that since backwards compatibility with the current implementation has
> > to be provided, the kernel has to continue supporting simultaneously both
> > the generic kernel -ERRORCODE return value ABI and the powerpc sc syscall
> > return ABI at least for PTRACE_EVENT_SECCOMP tracers. Consequently, since
> > the point of __secure_computing() invocation and up to the point of
> > conversion in syscall_exit_prepare(), gpr[3] may be set according to either
> > of these two ABIs. An attempt to address code inconsistencies in syscall
> > error return handling that were introduced as a side effect of the dual
> > ABI support follows in a separate patch.
>
> What do you mean by "backwards compatibility" here ? backwards
> compatibility applies only to userspace API doesn't it ? So if there was
> no way to trigger the problem previously, what does it mean ?
ptrace and seccomp users are pretty much userspace, aren't they?
> >
> > Link:
> > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.kernel.org%2Fdoc%2Fhtml%2Flatest%2Farch%2Fpowerpc%2Fsyscall64-abi.html%23return-value&data=05%7C02%7Cchristophe.leroy%40csgroup.eu%7Cc2cf590281c24fe1478408dd3efe4a3e%7C8b87af7d86474dc78df45f69a2011bb5%7C0%7C0%7C638735984085033893%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=EEP9s6k%2Fs5VfqWgrs6VXi879HEfJ8BYOOJ8InmmVTQA%3D&reserved=0
> > [1]
> > Signed-off-by: Dmitry V. Levin
> > Reviewed-by: Alexey Gladkov
> > ---
> > arch/powerpc/include/asm/syscall.h | 6 +-
> > 1 file changed, 5 insertions(+), 1 deletion(-)
> >
> > diff --git a/arch/powerpc/include/asm/syscall.h
> > b/arch/powerpc/include/asm/syscall.h
> > index 3dd36c5e334a..422d7735ace6 100644
> > --- a/arch/powerpc/include/asm/syscall.h
> > +++ b/arch/powerpc/include/asm/syscall.h
> > @@ -82,7 +82,11 @@ static inline void syscall_set_return_value(struct
> > task_struct *task,
> >*/
> > if (error) {
> > regs->ccr |= 0x1000L;
> > - regs->gpr[3] = error;
> > + /*
> > + * In case of an error regs->gpr[3] contains
> > + * a positive ERRORCODE.
> > + */
> > + regs->gpr[3] = -error;
> > } else {
> > regs->ccr &= ~0x1000L;
> > regs->gpr[3] = val;
>
--
Eugene Syromyatnikov
mailto:evg...@gmail.com
xmpp:esyr@jabber.{ru|org}
