Re: 5.11 regression: "ia64: add support for TIF_NOTIFY_SIGNAL" breaks ia64 boot

[Jens Axboe <axboe@kernel.dk>]

On 2/22/21 5:34 PM, Jens Axboe wrote:
> On 2/22/21 4:53 PM, Sergei Trofimovich wrote:
>> On Mon, 22 Feb 2021 16:34:50 -0700
>> Jens Axboe <axboe@kernel.dk> wrote:
>>
>>> On 2/22/21 4:05 PM, Sergei Trofimovich wrote:
>>>> Hia Jens!
>>>>
>>>> Tried 5.11 on rx3600 box and noticed it has
>>>> a problem handling init (5.10 booted fine):
>>>>
>>>> INIT: version 2.98 booting
>>>>
>>>>    OpenRC 0.42.1 is starting up Gentoo Linux (ia64)
>>>>
>>>> mkdir `/run/openrc': Read-only file system
>>>> mkdir `/run/openrc/starting': No such file or directory
>>>> mkdir `/run/openrc/started': No such file or directory
>>>> mkdir `/run/openrc/stopping': No such file or directory
>>>> mkdir `/run/openrc/inactive': No such file or directory
>>>> mkdir `/run/openrc/wasinactive': No such file or directory
>>>> mkdir `/run/openrc/failed': No such file or directory
>>>> mkdir `/run/openrc/hotplugged': No such file or directory
>>>> mkdir `/run/openrc/daemons': No such file or directory
>>>> mkdir `/run[   14.595059] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
>>>> [   14.599059] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b ]---
>>>>
>>>> I suspect we build bad signal stack frame for userspace.
>>>>
>>>> With a bit of #define DEBUG_SIG 1 enabled the signals are SIGCHLD:
>>>>
>>>> [   34.969771] SIG deliver (gendepends.sh:69): sig=17 sp=60000fffff6aeaa0 ip=a000000000040740 handler=000000004b4c59b6
>>>> [   34.969948] SIG deliver (init:1): sig=17 sp=60000fffff1ccc50 ip=a000000000040740 handler=000000004638b9e5
>>>> [   34.969948] SIG deliver (gendepends.sh:69): sig=17 sp=60000fffff6adf90 ip=a000000000040740 handler=000000004b4c59b6
>>>> [   34.973948] SIG deliver (init:1): sig=17 sp=60000fffff1cc140 ip=a000000000040740 handler=000000004638b9e5
>>>> [   34.973948] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
>>>> [   34.973948] SIG deliver (gendepends.sh:69): sig=17 sp=60000fffff6ad480 ip=a000000000040740 handler=000000004b4c59b6
>>>> [   34.973948] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b ]---
>>>>
>>>> Bisect points at:
>>>>
>>>> commit b269c229b0e89aedb7943c06673b56b6052cf5e5
>>>> Author: Jens Axboe <axboe@kernel.dk>
>>>> Date:   Fri Oct 9 14:49:43 2020 -0600
>>>>
>>>>     ia64: add support for TIF_NOTIFY_SIGNAL
>>>>
>>>>     Wire up TIF_NOTIFY_SIGNAL handling for ia64.
>>>>
>>>>     Cc: linux-ia64@vger.kernel.org
>>>>     [axboe: added fixes from Mike Rapoport <rppt@kernel.org>]
>>>>     Signed-off-by: Jens Axboe <axboe@kernel.dk>
>>>>
>>>> diff --git a/arch/ia64/include/asm/thread_info.h b/arch/ia64/include/asm/thread_info.h
>>>> index 64a1011f6812..51d20cb37706 100644
>>>> --- a/arch/ia64/include/asm/thread_info.h
>>>> +++ b/arch/ia64/include/asm/thread_info.h
>>>> @@ -103,6 +103,7 @@ struct thread_info {
>>>>  #define TIF_SYSCALL_TRACE      2       /* syscall trace active */
>>>>  #define TIF_SYSCALL_AUDIT      3       /* syscall auditing active */
>>>>  #define TIF_SINGLESTEP         4       /* restore singlestep on return to user mode */
>>>> +#define TIF_NOTIFY_SIGNAL      5       /* signal notification exist */
>>>>  #define TIF_NOTIFY_RESUME      6       /* resumption notification requested */
>>>>  #define TIF_MEMDIE             17      /* is terminating due to OOM killer */
>>>>  #define TIF_MCA_INIT           18      /* this task is processing MCA or INIT */
>>>> @@ -115,6 +116,7 @@ struct thread_info {
>>>>  #define _TIF_SINGLESTEP                (1 << TIF_SINGLESTEP)
>>>>  #define _TIF_SYSCALL_TRACEAUDIT        (_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP)
>>>>  #define _TIF_NOTIFY_RESUME     (1 << TIF_NOTIFY_RESUME)
>>>> +#define _TIF_NOTIFY_SIGNAL     (1 << TIF_NOTIFY_SIGNAL)
>>>>  #define _TIF_SIGPENDING                (1 << TIF_SIGPENDING)
>>>>  #define _TIF_NEED_RESCHED      (1 << TIF_NEED_RESCHED)
>>>>  #define _TIF_MCA_INIT          (1 << TIF_MCA_INIT)
>>>> @@ -124,7 +126,7 @@ struct thread_info {
>>>>
>>>>  /* "work to do on user-return" bits */
>>>>  #define TIF_ALLWORK_MASK       (_TIF_SIGPENDING|_TIF_NOTIFY_RESUME|_TIF_SYSCALL_AUDIT|\
>>>> -                                _TIF_NEED_RESCHED|_TIF_SYSCALL_TRACE)
>>>> +                                _TIF_NEED_RESCHED|_TIF_SYSCALL_TRACE|_TIF_NOTIFY_SIGNAL)
>>>>  /* like TIF_ALLWORK_BITS but sans TIF_SYSCALL_TRACE or TIF_SYSCALL_AUDIT */
>>>>  #define TIF_WORK_MASK          (TIF_ALLWORK_MASK&~(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT))
>>>>
>>>> diff --git a/arch/ia64/kernel/process.c b/arch/ia64/kernel/process.c
>>>> index 6b61a703bcf5..8d4e1cab9190 100644
>>>> --- a/arch/ia64/kernel/process.c
>>>> +++ b/arch/ia64/kernel/process.c
>>>> @@ -171,7 +171,8 @@ do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
>>>>         }
>>>>
>>>>         /* deal with pending signal delivery */
>>>> -       if (test_thread_flag(TIF_SIGPENDING)) {
>>>> +       if (test_thread_flag(TIF_SIGPENDING) ||
>>>> +           test_thread_flag(TIF_NOTIFY_SIGNAL)) {
>>>>                 local_irq_enable();     /* force interrupt enable */
>>>>                 ia64_do_signal(scr, in_syscall);
>>>>
>>>> which looks benign, but it enables a bit of conditional
>>>> TIF_NOTIFY_SIGNAL handling I don't understand.
>>>>
>>>> Can you help me get what is the interaction between
>>>> TIF_NOTIFY_SIGNAL and TIF_SIGPENDING for
>>>> simple processes without io_uring use case?
>>>>
>>>> I wonder if it's ia64_do_signal()' generates a signal
>>>> delivery when it should not.  
>>>
>>> Can you test:
>>>
>>> https://marc.info/?l=linux-ia64&m=161187407609443&w=1
>>>
>>> with the addition mentioned here:
>>
>> Not enough:
>>
>> mkdir `/run/openrc': Read-only file system
>> mkdir `/run/openrc/starting': No such file or directory
>> mkdir `/run/openrc/started': No such file or directory
>> mkdir `/run/openrc/stopping': No such file or directory
>> mkdir `/run/openrc/inactive': No such file or directory
>> mkdir `/run/openrc/wasinactive': No such file or directory
>> mkdir `/run/openrc/failed': No such file or directory
>> mkdir `/run/openrc/hotplugged': No such file or directory
>> mkdir `/run/openrc/daemons': No such file or directory
>> [   14.554357] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
>> [   14.554357] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b ]---
>> mkdir `/run/openrc/options': No such file or directory
>> mkdir `/run/openrc/exclusive': No such file or directory
>> mkdir `/run/openrc/scheduled': No such file or directory
>> mkdir `/run/openrc/tmp': No such file or directory
>>
>>> https://marc.info/?l=linux-ia64&m=161187470709706&w=1
>>>
>>> if needed?
>>
>> Two patches above do fix the boot \o/ But have a lot of spam about
>> 'signal 0' delivery to a bunch of processes:
>>
>>  * Mounting /proc ...
>>  [ ok ]
>>  * Mounting /run ...
>>  * /run/openrc: creating directory
>>  * /run/lock: creating directory
>>  * /run/lock: correcting owner
>>  * Caching service dependencies ...
>>  [ ok ]
>>  * Mounting /sys ...
>>  [ ok ]
>>  * Mounting debug filesystem ...
>>  [ ok ]
>>  * Mounting efivarfs filesystem ...
>>  [ ok ]
>>  * sysfs: caught unknown signal 0
>>  * openrc: caught unknown signal 0
>>  * Mounting cgroup filesystem ...
> 
> That's an improvement! Let me take a look at this tonight and see if I
> can figure out what's going on. But yes, it's the ia64 signal delivery
> being just different enough from the norm that it apparently triggers
> some weirdness.

Is this any better?


diff --git a/arch/ia64/kernel/signal.c b/arch/ia64/kernel/signal.c
index e67b22fc3c60..11891240aa5c 100644
--- a/arch/ia64/kernel/signal.c
+++ b/arch/ia64/kernel/signal.c
@@ -340,8 +340,10 @@ ia64_do_signal (struct sigscratch *scr, long in_syscall)
 	 * This only loops in the rare cases of handle_signal() failing, in which case we
 	 * need to push through a forced SIGSEGV.
 	 */
-	while (1) {
-		get_signal(&ksig);
+	do {
+		ksig.sig = 0;
+		if (!get_signal(&ksig))
+			break;
 
 		/*
 		 * get_signal() may have run a debugger (via notify_parent())
@@ -358,9 +360,6 @@ ia64_do_signal (struct sigscratch *scr, long in_syscall)
 			 */
 			restart = 0;
 
-		if (ksig.sig <= 0)
-			break;
-
 		if (unlikely(restart)) {
 			switch (errno) {
 			case ERESTART_RESTARTBLOCK:
@@ -387,7 +386,7 @@ ia64_do_signal (struct sigscratch *scr, long in_syscall)
 		 */
 		if (handle_signal(&ksig, scr))
 			return;
-	}
+	} while (ksig.sig > 0);
 
 	/* Did we come from a system call? */
 	if (restart) {

-- 
Jens Axboe