| Message ID | 1450303792-27470-1-git-send-email-mathieu.desnoyers@efficios.com (mailing list archive) |
|---|---|
| State | Accepted, archived |
| Headers | show |
----- On Dec 16, 2015, at 5:09 PM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote: > When testing liburcu on a 3.18 Linux kernel, 2-core MIPS (cpu model : > Ingenic JZRISC V4.15 FPU V0.0), we notice that a blocked sys_futex > FUTEX_WAIT returns -1, errno=ENOSYS when interrupted by a SA_RESTART > signal handler. This spurious ENOSYS behavior causes hangs in liburcu > 0.9.x. Running a MIPS 3.18 kernel under a QEMU emulator exhibits the > same behavior. This might affect earlier kernels. > > This issue appears to be fixed in 3.18.y stable kernels and 3.19, but > nevertheless, we should try to handle this kernel bug more gracefully > than a user-space hang due to unexpected spurious ENOSYS return value. It's actually fixed in 3.19, but not in 3.18.y stable kernels. The Linux kernel upstream fix commit is: e967ef02 "MIPS: Fix restart of indirect syscalls" I've created a small test program that could also be used on parisc to check if it suffers from the same issue (see attached). On bogus mips kernels, we see the following output: [OK] Test program with pid: 5748 SIGUSR1 handler [FAIL] futex returns -1, Function not implemented Let me know if someone can try it out on a parisc kernel. Thanks! Mathieu > > Therefore, fallback on the "async-safe" version of compat_futex in those > situations where FUTEX_WAIT returns ENOSYS. This async-safe fallback has > the nice property of being OK to use concurrently with other FUTEX_WAKE > and FUTEX_WAIT futex() calls, because it's simply a busy-wait scheme. > > We suspect that parisc might be affected by a similar issue (Debian > build bots reported a similar hang on both mips and parisc), but we do > not have access to the hardware required to test this hypothesis. > > Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> > CC: Michael Jeanson <mjeanson@efficios.com> > CC: Paul E. McKenney <paulmck@linux.vnet.ibm.com> > CC: Ralf Baechle <ralf@linux-mips.org> > CC: linux-mips@linux-mips.org > CC: linux-kernel@vger.kernel.org > CC: "James E.J. Bottomley" <jejb@parisc-linux.org> > CC: Helge Deller <deller@gmx.de> > CC: linux-parisc@vger.kernel.org > --- > compat_futex.c | 2 ++ > urcu/futex.h | 12 +++++++++++- > 2 files changed, 13 insertions(+), 1 deletion(-) > > diff --git a/compat_futex.c b/compat_futex.c > index b7f78f0..9e918fe 100644 > --- a/compat_futex.c > +++ b/compat_futex.c > @@ -111,6 +111,8 @@ end: > * _ASYNC SIGNAL-SAFE_. > * For now, timeout, uaddr2 and val3 are unused. > * Waiter will busy-loop trying to read the condition. > + * It is OK to use compat_futex_async() on a futex address on which > + * futex() WAKE operations are also performed. > */ > > int compat_futex_async(int32_t *uaddr, int op, int32_t val, > diff --git a/urcu/futex.h b/urcu/futex.h > index 4d16cfa..a17eda8 100644 > --- a/urcu/futex.h > +++ b/urcu/futex.h > @@ -73,7 +73,17 @@ static inline int futex_noasync(int32_t *uaddr, int op, > int32_t val, > > ret = futex(uaddr, op, val, timeout, uaddr2, val3); > if (caa_unlikely(ret < 0 && errno == ENOSYS)) { > - return compat_futex_noasync(uaddr, op, val, timeout, > + /* > + * The fallback on ENOSYS is the async-safe version of > + * the compat futex implementation, because the > + * async-safe compat implementation allows being used > + * concurrently with calls to futex(). Indeed, sys_futex > + * FUTEX_WAIT, on some architectures (e.g. mips), within > + * a given process, spuriously return ENOSYS due to > + * signal restart bugs on some kernel versions (e.g. > + * Linux kernel 3.18 and possibly earlier). > + */ > + return compat_futex_async(uaddr, op, val, timeout, > uaddr2, val3); > } > return ret; > -- > 2.1.4
I believe e967ef02 "MIPS: Fix restart of indirect syscalls" should be backported to all stable kernels. It would be a surprising coincidence if parisc suffers from the same problem. --Ed On Thu, Dec 17, 2015 at 4:54 AM, Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote: > ----- On Dec 16, 2015, at 5:09 PM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote: > >> When testing liburcu on a 3.18 Linux kernel, 2-core MIPS (cpu model : >> Ingenic JZRISC V4.15 FPU V0.0), we notice that a blocked sys_futex >> FUTEX_WAIT returns -1, errno=ENOSYS when interrupted by a SA_RESTART >> signal handler. This spurious ENOSYS behavior causes hangs in liburcu >> 0.9.x. Running a MIPS 3.18 kernel under a QEMU emulator exhibits the >> same behavior. This might affect earlier kernels. >> >> This issue appears to be fixed in 3.18.y stable kernels and 3.19, but >> nevertheless, we should try to handle this kernel bug more gracefully >> than a user-space hang due to unexpected spurious ENOSYS return value. > > It's actually fixed in 3.19, but not in 3.18.y stable kernels. The > Linux kernel upstream fix commit is: > e967ef02 "MIPS: Fix restart of indirect syscalls" > > I've created a small test program that could also be used on parisc > to check if it suffers from the same issue (see attached). > > On bogus mips kernels, we see the following output: > [OK] Test program with pid: 5748 SIGUSR1 handler > [FAIL] futex returns -1, Function not implemented > > Let me know if someone can try it out on a parisc kernel. > > Thanks! > > Mathieu > >> >> Therefore, fallback on the "async-safe" version of compat_futex in those >> situations where FUTEX_WAIT returns ENOSYS. This async-safe fallback has >> the nice property of being OK to use concurrently with other FUTEX_WAKE >> and FUTEX_WAIT futex() calls, because it's simply a busy-wait scheme. >> >> We suspect that parisc might be affected by a similar issue (Debian >> build bots reported a similar hang on both mips and parisc), but we do >> not have access to the hardware required to test this hypothesis. >> >> Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> >> CC: Michael Jeanson <mjeanson@efficios.com> >> CC: Paul E. McKenney <paulmck@linux.vnet.ibm.com> >> CC: Ralf Baechle <ralf@linux-mips.org> >> CC: linux-mips@linux-mips.org >> CC: linux-kernel@vger.kernel.org >> CC: "James E.J. Bottomley" <jejb@parisc-linux.org> >> CC: Helge Deller <deller@gmx.de> >> CC: linux-parisc@vger.kernel.org >> --- >> compat_futex.c | 2 ++ >> urcu/futex.h | 12 +++++++++++- >> 2 files changed, 13 insertions(+), 1 deletion(-) >> >> diff --git a/compat_futex.c b/compat_futex.c >> index b7f78f0..9e918fe 100644 >> --- a/compat_futex.c >> +++ b/compat_futex.c >> @@ -111,6 +111,8 @@ end: >> * _ASYNC SIGNAL-SAFE_. >> * For now, timeout, uaddr2 and val3 are unused. >> * Waiter will busy-loop trying to read the condition. >> + * It is OK to use compat_futex_async() on a futex address on which >> + * futex() WAKE operations are also performed. >> */ >> >> int compat_futex_async(int32_t *uaddr, int op, int32_t val, >> diff --git a/urcu/futex.h b/urcu/futex.h >> index 4d16cfa..a17eda8 100644 >> --- a/urcu/futex.h >> +++ b/urcu/futex.h >> @@ -73,7 +73,17 @@ static inline int futex_noasync(int32_t *uaddr, int op, >> int32_t val, >> >> ret = futex(uaddr, op, val, timeout, uaddr2, val3); >> if (caa_unlikely(ret < 0 && errno == ENOSYS)) { >> - return compat_futex_noasync(uaddr, op, val, timeout, >> + /* >> + * The fallback on ENOSYS is the async-safe version of >> + * the compat futex implementation, because the >> + * async-safe compat implementation allows being used >> + * concurrently with calls to futex(). Indeed, sys_futex >> + * FUTEX_WAIT, on some architectures (e.g. mips), within >> + * a given process, spuriously return ENOSYS due to >> + * signal restart bugs on some kernel versions (e.g. >> + * Linux kernel 3.18 and possibly earlier). >> + */ >> + return compat_futex_async(uaddr, op, val, timeout, >> uaddr2, val3); >> } >> return ret; >> -- >> 2.1.4 > > -- > Mathieu Desnoyers > EfficiOS Inc. > http://www.efficios.com -- To unsubscribe from this list: send the line "unsubscribe linux-parisc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hello Mathieu, > > When testing liburcu on a 3.18 Linux kernel, 2-core MIPS (cpu model : > > Ingenic JZRISC V4.15 FPU V0.0), we notice that a blocked sys_futex > > FUTEX_WAIT returns -1, errno=ENOSYS when interrupted by a SA_RESTART > > signal handler. This spurious ENOSYS behavior causes hangs in liburcu > > 0.9.x. Running a MIPS 3.18 kernel under a QEMU emulator exhibits the > > same behavior. This might affect earlier kernels. > > > > This issue appears to be fixed in 3.18.y stable kernels and 3.19, but > > nevertheless, we should try to handle this kernel bug more gracefully > > than a user-space hang due to unexpected spurious ENOSYS return value. > > It's actually fixed in 3.19, but not in 3.18.y stable kernels. The > Linux kernel upstream fix commit is: > e967ef02 "MIPS: Fix restart of indirect syscalls" But that patch fixes mips only. > I've created a small test program that could also be used on parisc > to check if it suffers from the same issue (see attached). > > On bogus mips kernels, we see the following output: > [OK] Test program with pid: 5748 SIGUSR1 handler > [FAIL] futex returns -1, Function not implemented I tested it on a recent 4.2 kernel on parisc. It fails as you describe: Testing futex sigrestart. Stop with CTRL-c. [OK] Test program with pid: 1361 SIGUSR1 handler [OK] Test program with pid: 1361 SIGUSR1 handler [FAIL] futex returns -1, Function not implemented [OK] Test program with pid: 1361 SIGUSR1 handler [FAIL] futex returns -1, Function not implemented strace gives: [pid 1329] futex(0x1210c, FUTEX_WAIT, -1, NULL <unfinished ...> [pid 1328] nanosleep({1, 0}, <unfinished ...> [pid 1329] <... futex resumed> ) = ? ERESTARTSYS (To be restarted if SA_RESTART is set) [pid 1329] write(2, "[FAIL] futex returns -1, Functio"..., 50[FAIL] futex returns -1, Function not implemented) > > Therefore, fallback on the "async-safe" version of compat_futex in those > > situations where FUTEX_WAIT returns ENOSYS. This async-safe fallback has > > the nice property of being OK to use concurrently with other FUTEX_WAKE > > and FUTEX_WAIT futex() calls, because it's simply a busy-wait scheme. > > > > We suspect that parisc might be affected by a similar issue (Debian > > build bots reported a similar hang on both mips and parisc), but we do > > not have access to the hardware required to test this hypothesis. If you want access to a machine, let me know. I'll try the patch below as well.. > > Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> > > CC: Michael Jeanson <mjeanson@efficios.com> > > CC: Paul E. McKenney <paulmck@linux.vnet.ibm.com> > > CC: Ralf Baechle <ralf@linux-mips.org> > > CC: linux-mips@linux-mips.org > > CC: linux-kernel@vger.kernel.org > > CC: "James E.J. Bottomley" <jejb@parisc-linux.org> > > CC: Helge Deller <deller@gmx.de> > > CC: linux-parisc@vger.kernel.org > > --- > > compat_futex.c | 2 ++ > > urcu/futex.h | 12 +++++++++++- > > 2 files changed, 13 insertions(+), 1 deletion(-) > > > > diff --git a/compat_futex.c b/compat_futex.c > > index b7f78f0..9e918fe 100644 > > --- a/compat_futex.c > > +++ b/compat_futex.c > > @@ -111,6 +111,8 @@ end: > > * _ASYNC SIGNAL-SAFE_. > > * For now, timeout, uaddr2 and val3 are unused. > > * Waiter will busy-loop trying to read the condition. > > + * It is OK to use compat_futex_async() on a futex address on which > > + * futex() WAKE operations are also performed. > > */ > > > > int compat_futex_async(int32_t *uaddr, int op, int32_t val, > > diff --git a/urcu/futex.h b/urcu/futex.h > > index 4d16cfa..a17eda8 100644 > > --- a/urcu/futex.h > > +++ b/urcu/futex.h > > @@ -73,7 +73,17 @@ static inline int futex_noasync(int32_t *uaddr, int op, > > int32_t val, > > > > ret = futex(uaddr, op, val, timeout, uaddr2, val3); > > if (caa_unlikely(ret < 0 && errno == ENOSYS)) { > > - return compat_futex_noasync(uaddr, op, val, timeout, > > + /* > > + * The fallback on ENOSYS is the async-safe version of > > + * the compat futex implementation, because the > > + * async-safe compat implementation allows being used > > + * concurrently with calls to futex(). Indeed, sys_futex > > + * FUTEX_WAIT, on some architectures (e.g. mips), within > > + * a given process, spuriously return ENOSYS due to > > + * signal restart bugs on some kernel versions (e.g. > > + * Linux kernel 3.18 and possibly earlier). > > + */ > > + return compat_futex_async(uaddr, op, val, timeout, > > uaddr2, val3); > > } > > return ret; -- To unsubscribe from this list: send the line "unsubscribe linux-parisc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
----- On Dec 17, 2015, at 11:22 AM, Helge Deller deller@gmx.de wrote: > Hello Mathieu, > >> > When testing liburcu on a 3.18 Linux kernel, 2-core MIPS (cpu model : >> > Ingenic JZRISC V4.15 FPU V0.0), we notice that a blocked sys_futex >> > FUTEX_WAIT returns -1, errno=ENOSYS when interrupted by a SA_RESTART >> > signal handler. This spurious ENOSYS behavior causes hangs in liburcu >> > 0.9.x. Running a MIPS 3.18 kernel under a QEMU emulator exhibits the >> > same behavior. This might affect earlier kernels. >> > >> > This issue appears to be fixed in 3.18.y stable kernels and 3.19, but >> > nevertheless, we should try to handle this kernel bug more gracefully >> > than a user-space hang due to unexpected spurious ENOSYS return value. >> >> It's actually fixed in 3.19, but not in 3.18.y stable kernels. The >> Linux kernel upstream fix commit is: >> e967ef02 "MIPS: Fix restart of indirect syscalls" > > But that patch fixes mips only. Indeed, I do not expect this commit to have any effect on parisc. > >> I've created a small test program that could also be used on parisc >> to check if it suffers from the same issue (see attached). >> >> On bogus mips kernels, we see the following output: >> [OK] Test program with pid: 5748 SIGUSR1 handler >> [FAIL] futex returns -1, Function not implemented > > I tested it on a recent 4.2 kernel on parisc. > It fails as you describe: > > Testing futex sigrestart. Stop with CTRL-c. > [OK] Test program with pid: 1361 SIGUSR1 handler > [OK] Test program with pid: 1361 SIGUSR1 handler > [FAIL] futex returns -1, Function not implemented > [OK] Test program with pid: 1361 SIGUSR1 handler > [FAIL] futex returns -1, Function not implemented > > strace gives: > [pid 1329] futex(0x1210c, FUTEX_WAIT, -1, NULL <unfinished ...> > [pid 1328] nanosleep({1, 0}, <unfinished ...> > [pid 1329] <... futex resumed> ) = ? ERESTARTSYS (To be restarted if > SA_RESTART is set) > [pid 1329] write(2, "[FAIL] futex returns -1, Functio"..., 50[FAIL] futex > returns -1, Function not implemented) Looks like parisc has an issue very similar to the one that has been fixed on MIPS by e967ef02 "MIPS: Fix restart of indirect syscalls". > > >> > Therefore, fallback on the "async-safe" version of compat_futex in those >> > situations where FUTEX_WAIT returns ENOSYS. This async-safe fallback has >> > the nice property of being OK to use concurrently with other FUTEX_WAKE >> > and FUTEX_WAIT futex() calls, because it's simply a busy-wait scheme. >> > >> > We suspect that parisc might be affected by a similar issue (Debian >> > build bots reported a similar hang on both mips and parisc), but we do >> > not have access to the hardware required to test this hypothesis. > > If you want access to a machine, let me know. > I'll try the patch below as well.. This would be very useful indeed, just to make sure our approach to futex fallback in liburcu works fine on parisc. I'm no parisc assembly expert though, but I suspect the issue would be quite similar to the one already fixed on MIPS. The existing fix for MIPS would be a good starting point to see if something similar is missing on parisc. When time allows, we should consider cleaning up my test case for restart of indirect system calls and add it to kselftest. It's the second architecture that has the same defect, which means this behavior is seldom tested. Thanks, Mathieu > >> > Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> >> > CC: Michael Jeanson <mjeanson@efficios.com> >> > CC: Paul E. McKenney <paulmck@linux.vnet.ibm.com> >> > CC: Ralf Baechle <ralf@linux-mips.org> >> > CC: linux-mips@linux-mips.org >> > CC: linux-kernel@vger.kernel.org >> > CC: "James E.J. Bottomley" <jejb@parisc-linux.org> >> > CC: Helge Deller <deller@gmx.de> >> > CC: linux-parisc@vger.kernel.org >> > --- >> > compat_futex.c | 2 ++ >> > urcu/futex.h | 12 +++++++++++- >> > 2 files changed, 13 insertions(+), 1 deletion(-) >> > >> > diff --git a/compat_futex.c b/compat_futex.c >> > index b7f78f0..9e918fe 100644 >> > --- a/compat_futex.c >> > +++ b/compat_futex.c >> > @@ -111,6 +111,8 @@ end: >> > * _ASYNC SIGNAL-SAFE_. >> > * For now, timeout, uaddr2 and val3 are unused. >> > * Waiter will busy-loop trying to read the condition. >> > + * It is OK to use compat_futex_async() on a futex address on which >> > + * futex() WAKE operations are also performed. >> > */ >> > >> > int compat_futex_async(int32_t *uaddr, int op, int32_t val, >> > diff --git a/urcu/futex.h b/urcu/futex.h >> > index 4d16cfa..a17eda8 100644 >> > --- a/urcu/futex.h >> > +++ b/urcu/futex.h >> > @@ -73,7 +73,17 @@ static inline int futex_noasync(int32_t *uaddr, int op, >> > int32_t val, >> > >> > ret = futex(uaddr, op, val, timeout, uaddr2, val3); >> > if (caa_unlikely(ret < 0 && errno == ENOSYS)) { >> > - return compat_futex_noasync(uaddr, op, val, timeout, >> > + /* >> > + * The fallback on ENOSYS is the async-safe version of >> > + * the compat futex implementation, because the >> > + * async-safe compat implementation allows being used >> > + * concurrently with calls to futex(). Indeed, sys_futex >> > + * FUTEX_WAIT, on some architectures (e.g. mips), within >> > + * a given process, spuriously return ENOSYS due to >> > + * signal restart bugs on some kernel versions (e.g. >> > + * Linux kernel 3.18 and possibly earlier). >> > + */ >> > + return compat_futex_async(uaddr, op, val, timeout, >> > uaddr2, val3); >> > } > > > return ret;
Hi Mathieu, On 18.12.2015 20:58, Mathieu Desnoyers wrote: > ----- On Dec 17, 2015, at 11:22 AM, Helge Deller deller@gmx.de wrote: > >> Hello Mathieu, >> >>>> When testing liburcu on a 3.18 Linux kernel, 2-core MIPS (cpu model : >>>> Ingenic JZRISC V4.15 FPU V0.0), we notice that a blocked sys_futex >>>> FUTEX_WAIT returns -1, errno=ENOSYS when interrupted by a SA_RESTART >>>> signal handler. This spurious ENOSYS behavior causes hangs in liburcu >>>> 0.9.x. Running a MIPS 3.18 kernel under a QEMU emulator exhibits the >>>> same behavior. This might affect earlier kernels. >>>> >>>> This issue appears to be fixed in 3.18.y stable kernels and 3.19, but >>>> nevertheless, we should try to handle this kernel bug more gracefully >>>> than a user-space hang due to unexpected spurious ENOSYS return value. >>> >>> It's actually fixed in 3.19, but not in 3.18.y stable kernels. The >>> Linux kernel upstream fix commit is: >>> e967ef02 "MIPS: Fix restart of indirect syscalls" >> >> But that patch fixes mips only. > > Indeed, I do not expect this commit to have any effect on parisc. > >> >>> I've created a small test program that could also be used on parisc >>> to check if it suffers from the same issue (see attached). >>> >>> On bogus mips kernels, we see the following output: >>> [OK] Test program with pid: 5748 SIGUSR1 handler >>> [FAIL] futex returns -1, Function not implemented >> >> I tested it on a recent 4.2 kernel on parisc. >> It fails as you describe: >> >> Testing futex sigrestart. Stop with CTRL-c. >> [OK] Test program with pid: 1361 SIGUSR1 handler >> [OK] Test program with pid: 1361 SIGUSR1 handler >> [FAIL] futex returns -1, Function not implemented >> [OK] Test program with pid: 1361 SIGUSR1 handler >> [FAIL] futex returns -1, Function not implemented >> >> strace gives: >> [pid 1329] futex(0x1210c, FUTEX_WAIT, -1, NULL <unfinished ...> >> [pid 1328] nanosleep({1, 0}, <unfinished ...> >> [pid 1329] <... futex resumed> ) = ? ERESTARTSYS (To be restarted if >> SA_RESTART is set) >> [pid 1329] write(2, "[FAIL] futex returns -1, Functio"..., 50[FAIL] futex >> returns -1, Function not implemented) > > Looks like parisc has an issue very similar to the one that > has been fixed on MIPS by e967ef02 "MIPS: Fix restart of indirect syscalls". Yes. >>>> Therefore, fallback on the "async-safe" version of compat_futex in those >>>> situations where FUTEX_WAIT returns ENOSYS. This async-safe fallback has >>>> the nice property of being OK to use concurrently with other FUTEX_WAKE >>>> and FUTEX_WAIT futex() calls, because it's simply a busy-wait scheme. >>>> >>>> We suspect that parisc might be affected by a similar issue (Debian >>>> build bots reported a similar hang on both mips and parisc), but we do >>>> not have access to the hardware required to test this hypothesis. >> >> If you want access to a machine, let me know. >> I'll try the patch below as well.. > > This would be very useful indeed, just to make sure our approach to > futex fallback in liburcu works fine on parisc. Yes, but will take me some time... > I'm no parisc assembly expert though, but I suspect the issue > would be quite similar to the one already fixed on MIPS. The > existing fix for MIPS would be a good starting point to see if > something similar is missing on parisc. Yes, I've already started to look into the parisc assembly parts. The problems seems to be both the same, the syscall number is not reserved during a syscall restart. We have problems with pthread cancellation in glibc too, maybe it's related to this bug. > When time allows, we should consider cleaning up my test case for > restart of indirect system calls and add it to kselftest. I was thinking of adding it to the Linux Test Project (LTP) :-) > It's > the second architecture that has the same defect, which means this > behavior is seldom tested. Helge -- To unsubscribe from this list: send the line "unsubscribe linux-parisc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Mathieu, On 18.12.2015 21:42, Helge Deller wrote: > On 18.12.2015 20:58, Mathieu Desnoyers wrote: >>>>> When testing liburcu on a 3.18 Linux kernel, 2-core MIPS (cpu model : >>>>> Ingenic JZRISC V4.15 FPU V0.0), we notice that a blocked sys_futex >>>>> FUTEX_WAIT returns -1, errno=ENOSYS when interrupted by a SA_RESTART >>>>> signal handler. This spurious ENOSYS behavior causes hangs in liburcu >>>>> 0.9.x. Running a MIPS 3.18 kernel under a QEMU emulator exhibits the >>>>> same behavior. This might affect earlier kernels. >>>>> >>>>> This issue appears to be fixed in 3.18.y stable kernels and 3.19, but >>>>> nevertheless, we should try to handle this kernel bug more gracefully >>>>> than a user-space hang due to unexpected spurious ENOSYS return value. >>>> >>>> It's actually fixed in 3.19, but not in 3.18.y stable kernels. The >>>> Linux kernel upstream fix commit is: >>>> e967ef02 "MIPS: Fix restart of indirect syscalls" >> Looks like parisc has an issue very similar to the one that >> has been fixed on MIPS by e967ef02 "MIPS: Fix restart of indirect syscalls". Yes, parisc is affected the same way. I've posted a patch to the parisc mailing list which fixes this issue for parisc and which I plan to push into stable kernels: http://thread.gmane.org/gmane.linux.ports.parisc/26243 Regarding your patch for liburcu: >>>>> Therefore, fallback on the "async-safe" version of compat_futex in those >>>>> situations where FUTEX_WAIT returns ENOSYS. This async-safe fallback has >>>>> the nice property of being OK to use concurrently with other FUTEX_WAKE >>>>> and FUTEX_WAIT futex() calls, because it's simply a busy-wait scheme. I've tested your patch. It does not produce any regressions on parisc, but I can't say for sure if it really works. ENOSYS is returned randomly, so maybe I didn't faced a situation where your patch actually was used. Helge -- To unsubscribe from this list: send the line "unsubscribe linux-parisc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
----- On Dec 19, 2015, at 5:37 AM, Helge Deller deller@gmx.de wrote: > Hi Mathieu, > > On 18.12.2015 21:42, Helge Deller wrote: >> On 18.12.2015 20:58, Mathieu Desnoyers wrote: >>>>>> When testing liburcu on a 3.18 Linux kernel, 2-core MIPS (cpu model : >>>>>> Ingenic JZRISC V4.15 FPU V0.0), we notice that a blocked sys_futex >>>>>> FUTEX_WAIT returns -1, errno=ENOSYS when interrupted by a SA_RESTART >>>>>> signal handler. This spurious ENOSYS behavior causes hangs in liburcu >>>>>> 0.9.x. Running a MIPS 3.18 kernel under a QEMU emulator exhibits the >>>>>> same behavior. This might affect earlier kernels. >>>>>> >>>>>> This issue appears to be fixed in 3.18.y stable kernels and 3.19, but >>>>>> nevertheless, we should try to handle this kernel bug more gracefully >>>>>> than a user-space hang due to unexpected spurious ENOSYS return value. >>>>> >>>>> It's actually fixed in 3.19, but not in 3.18.y stable kernels. The >>>>> Linux kernel upstream fix commit is: >>>>> e967ef02 "MIPS: Fix restart of indirect syscalls" > >>> Looks like parisc has an issue very similar to the one that >>> has been fixed on MIPS by e967ef02 "MIPS: Fix restart of indirect syscalls". > > Yes, parisc is affected the same way. > I've posted a patch to the parisc mailing list which fixes this issue for > parisc and which I plan to push into stable kernels: > http://thread.gmane.org/gmane.linux.ports.parisc/26243 > > Regarding your patch for liburcu: > >>>>>> Therefore, fallback on the "async-safe" version of compat_futex in those >>>>>> situations where FUTEX_WAIT returns ENOSYS. This async-safe fallback has >>>>>> the nice property of being OK to use concurrently with other FUTEX_WAKE >>>>>> and FUTEX_WAIT futex() calls, because it's simply a busy-wait scheme. > > I've tested your patch. It does not produce any regressions on parisc, but I > can't > say for sure if it really works. ENOSYS is returned randomly, so maybe I didn't > faced a situation where your patch actually was used. If you ran make check and make regtest, and nothing fails/hangs, you should be OK. liburcu runs very heavy stress-tests which makes it likely to hit race conditions repeatedly. Thanks! Mathieu > > Helge
On 20.12.2015 15:11, Mathieu Desnoyers wrote: > ----- On Dec 19, 2015, at 5:37 AM, Helge Deller deller@gmx.de wrote: > >> Hi Mathieu, >> >> On 18.12.2015 21:42, Helge Deller wrote: >>> On 18.12.2015 20:58, Mathieu Desnoyers wrote: >>>>>>> When testing liburcu on a 3.18 Linux kernel, 2-core MIPS (cpu model : >>>>>>> Ingenic JZRISC V4.15 FPU V0.0), we notice that a blocked sys_futex >>>>>>> FUTEX_WAIT returns -1, errno=ENOSYS when interrupted by a SA_RESTART >>>>>>> signal handler. This spurious ENOSYS behavior causes hangs in liburcu >>>>>>> 0.9.x. Running a MIPS 3.18 kernel under a QEMU emulator exhibits the >>>>>>> same behavior. This might affect earlier kernels. >>>>>>> >>>>>>> This issue appears to be fixed in 3.18.y stable kernels and 3.19, but >>>>>>> nevertheless, we should try to handle this kernel bug more gracefully >>>>>>> than a user-space hang due to unexpected spurious ENOSYS return value. >>>>>> >>>>>> It's actually fixed in 3.19, but not in 3.18.y stable kernels. The >>>>>> Linux kernel upstream fix commit is: >>>>>> e967ef02 "MIPS: Fix restart of indirect syscalls" >> >>>> Looks like parisc has an issue very similar to the one that >>>> has been fixed on MIPS by e967ef02 "MIPS: Fix restart of indirect syscalls". >> >> Yes, parisc is affected the same way. >> I've posted a patch to the parisc mailing list which fixes this issue for >> parisc and which I plan to push into stable kernels: >> http://thread.gmane.org/gmane.linux.ports.parisc/26243 >> >> Regarding your patch for liburcu: >> >>>>>>> Therefore, fallback on the "async-safe" version of compat_futex in those >>>>>>> situations where FUTEX_WAIT returns ENOSYS. This async-safe fallback has >>>>>>> the nice property of being OK to use concurrently with other FUTEX_WAKE >>>>>>> and FUTEX_WAIT futex() calls, because it's simply a busy-wait scheme. >> >> I've tested your patch. It does not produce any regressions on parisc, but I >> can't >> say for sure if it really works. ENOSYS is returned randomly, so maybe I didn't >> faced a situation where your patch actually was used. > > If you ran make check and make regtest, and nothing > fails/hangs, you should be OK. Yes, I did run both. > liburcu runs very heavy > stress-tests which makes it likely to hit race conditions > repeatedly. Helge -- To unsubscribe from this list: send the line "unsubscribe linux-parisc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/compat_futex.c b/compat_futex.c index b7f78f0..9e918fe 100644 --- a/compat_futex.c +++ b/compat_futex.c @@ -111,6 +111,8 @@ end: * _ASYNC SIGNAL-SAFE_. * For now, timeout, uaddr2 and val3 are unused. * Waiter will busy-loop trying to read the condition. + * It is OK to use compat_futex_async() on a futex address on which + * futex() WAKE operations are also performed. */ int compat_futex_async(int32_t *uaddr, int op, int32_t val, diff --git a/urcu/futex.h b/urcu/futex.h index 4d16cfa..a17eda8 100644 --- a/urcu/futex.h +++ b/urcu/futex.h @@ -73,7 +73,17 @@ static inline int futex_noasync(int32_t *uaddr, int op, int32_t val, ret = futex(uaddr, op, val, timeout, uaddr2, val3); if (caa_unlikely(ret < 0 && errno == ENOSYS)) { - return compat_futex_noasync(uaddr, op, val, timeout, + /* + * The fallback on ENOSYS is the async-safe version of + * the compat futex implementation, because the + * async-safe compat implementation allows being used + * concurrently with calls to futex(). Indeed, sys_futex + * FUTEX_WAIT, on some architectures (e.g. mips), within + * a given process, spuriously return ENOSYS due to + * signal restart bugs on some kernel versions (e.g. + * Linux kernel 3.18 and possibly earlier). + */ + return compat_futex_async(uaddr, op, val, timeout, uaddr2, val3); } return ret;
When testing liburcu on a 3.18 Linux kernel, 2-core MIPS (cpu model : Ingenic JZRISC V4.15 FPU V0.0), we notice that a blocked sys_futex FUTEX_WAIT returns -1, errno=ENOSYS when interrupted by a SA_RESTART signal handler. This spurious ENOSYS behavior causes hangs in liburcu 0.9.x. Running a MIPS 3.18 kernel under a QEMU emulator exhibits the same behavior. This might affect earlier kernels. This issue appears to be fixed in 3.18.y stable kernels and 3.19, but nevertheless, we should try to handle this kernel bug more gracefully than a user-space hang due to unexpected spurious ENOSYS return value. Therefore, fallback on the "async-safe" version of compat_futex in those situations where FUTEX_WAIT returns ENOSYS. This async-safe fallback has the nice property of being OK to use concurrently with other FUTEX_WAKE and FUTEX_WAIT futex() calls, because it's simply a busy-wait scheme. We suspect that parisc might be affected by a similar issue (Debian build bots reported a similar hang on both mips and parisc), but we do not have access to the hardware required to test this hypothesis. Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> CC: Michael Jeanson <mjeanson@efficios.com> CC: Paul E. McKenney <paulmck@linux.vnet.ibm.com> CC: Ralf Baechle <ralf@linux-mips.org> CC: linux-mips@linux-mips.org CC: linux-kernel@vger.kernel.org CC: "James E.J. Bottomley" <jejb@parisc-linux.org> CC: Helge Deller <deller@gmx.de> CC: linux-parisc@vger.kernel.org --- compat_futex.c | 2 ++ urcu/futex.h | 12 +++++++++++- 2 files changed, 13 insertions(+), 1 deletion(-)