| Message ID | 20240902195841.1455273-1-adhemerval.zanella@linaro.org |
|---|---|
| State | Accepted |
| Commit | 49c3682ce18fc283433b759ea606bdc26df6b829 |
| Headers | show |
| Series | linux: sparc: Fix syscall_cancel for LEON | expand |
On Mon, 2024-09-02 at 16:58 -0300, Adhemerval Zanella wrote: > LEON2/LEON3 are both sparcv8, which does not support branch hints > (bne,pn) nor the return instruction. > > Checked with a build for sparcv8-linux-gnu targetting leon. I also > check some cancellation tests with qemu-system (targeting LEON3). ^^^^^ This should be "checked". > --- > sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S | 9 +++++++++ > 1 file changed, 9 insertions(+) > > diff --git a/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S b/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S > index aa5c658ce1..0db93c77bf 100644 > --- a/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S > +++ b/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S > @@ -42,7 +42,11 @@ __syscall_cancel_arch_start: > __syscall_do_cancel() */ > ld [%i0], %g2 > andcc %g2, TCB_CANCELED_BITMASK, %g0 > +#ifdef __sparcv9 > bne,pn %icc, 2f > +#else > + bne 2f > +#endif > /* Issue a 6 argument syscall. */ > mov %i1, %g1 > mov %i2, %o0 > @@ -60,8 +64,13 @@ __syscall_cancel_arch_end: > sub %g0, %o0, %o0 > 1: > mov %o0, %i0 > +#ifdef __sparcv9 > return %i7+8 > nop > +#else > + jmp %i7+8 > + restore > +#endif > > 2: > call __syscall_do_cancel, 0 Acked-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de> Thanks, Adrian
On 26/09/24 15:07, John Paul Adrian Glaubitz wrote: > On Mon, 2024-09-02 at 16:58 -0300, Adhemerval Zanella wrote: >> LEON2/LEON3 are both sparcv8, which does not support branch hints >> (bne,pn) nor the return instruction. >> >> Checked with a build for sparcv8-linux-gnu targetting leon. I also >> check some cancellation tests with qemu-system (targeting LEON3). > ^^^^^ > This should be "checked". Ack. > >> --- >> sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S | 9 +++++++++ >> 1 file changed, 9 insertions(+) >> >> diff --git a/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S b/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S >> index aa5c658ce1..0db93c77bf 100644 >> --- a/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S >> +++ b/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S >> @@ -42,7 +42,11 @@ __syscall_cancel_arch_start: >> __syscall_do_cancel() */ >> ld [%i0], %g2 >> andcc %g2, TCB_CANCELED_BITMASK, %g0 >> +#ifdef __sparcv9 >> bne,pn %icc, 2f >> +#else >> + bne 2f >> +#endif >> /* Issue a 6 argument syscall. */ >> mov %i1, %g1 >> mov %i2, %o0 >> @@ -60,8 +64,13 @@ __syscall_cancel_arch_end: >> sub %g0, %o0, %o0 >> 1: >> mov %o0, %i0 >> +#ifdef __sparcv9 >> return %i7+8 >> nop >> +#else >> + jmp %i7+8 >> + restore >> +#endif >> >> 2: >> call __syscall_do_cancel, 0 > > Acked-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de> > Thanks, I will install it.
Hi Adhemerval, On Tue, 2024-10-01 at 10:36 -0300, Adhemerval Zanella Netto wrote: > > Acked-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de> > > > > Thanks, I will install it. Thanks a lot for committing both fixes. Do you also have any news regarding that restartable syscall bug on SPARC that was reported and where you came up with a patch? Adrian
On 01/10/24 11:07, John Paul Adrian Glaubitz wrote: > Hi Adhemerval, > > On Tue, 2024-10-01 at 10:36 -0300, Adhemerval Zanella Netto wrote: >>> Acked-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de> >>> >> >> Thanks, I will install it. > > Thanks a lot for committing both fixes. > > Do you also have any news regarding that restartable syscall bug on SPARC > that was reported and where you came up with a patch? Florian has acked the support patch for the testcase, but I did not hear anything from Rene (which was the one that originally complained about this issue). At least I haven't see any regression on glibc testsuite.
Adhemerval Zanella Netto <adhemerval.zanella@linaro.org> writes: > On 01/10/24 11:07, John Paul Adrian Glaubitz wrote: >> Hi Adhemerval, >> >> On Tue, 2024-10-01 at 10:36 -0300, Adhemerval Zanella Netto wrote: >>>> Acked-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de> >>>> >>> >>> Thanks, I will install it. >> >> Thanks a lot for committing both fixes. >> >> Do you also have any news regarding that restartable syscall bug on SPARC >> that was reported and where you came up with a patch? > > Florian has acked the support patch for the testcase, but I did not hear > anything from Rene (which was the one that originally complained about > this issue). At least I haven't see any regression on glibc testsuite. I can test it as I can reproduce the issue easily with job control in the shell.
diff --git a/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S b/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S index aa5c658ce1..0db93c77bf 100644 --- a/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S +++ b/sysdeps/unix/sysv/linux/sparc/sparc32/syscall_cancel.S @@ -42,7 +42,11 @@ __syscall_cancel_arch_start: __syscall_do_cancel() */ ld [%i0], %g2 andcc %g2, TCB_CANCELED_BITMASK, %g0 +#ifdef __sparcv9 bne,pn %icc, 2f +#else + bne 2f +#endif /* Issue a 6 argument syscall. */ mov %i1, %g1 mov %i2, %o0 @@ -60,8 +64,13 @@ __syscall_cancel_arch_end: sub %g0, %o0, %o0 1: mov %o0, %i0 +#ifdef __sparcv9 return %i7+8 nop +#else + jmp %i7+8 + restore +#endif 2: call __syscall_do_cancel, 0