This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] posix: Fix improper assert in Linux posix_spawn (BZ#22273)
On 17/10/2017 18:36, Florian Weimer wrote:
> * Adhemerval Zanella:
>
>> diff --git a/sysdeps/unix/sysv/linux/spawni.c b/sysdeps/unix/sysv/linux/spawni.c
>> index dea1650..d6acc1e 100644
>> --- a/sysdeps/unix/sysv/linux/spawni.c
>> +++ b/sysdeps/unix/sysv/linux/spawni.c
>> @@ -365,9 +365,17 @@ __spawnix (pid_t * pid, const char *file,
>> if (new_pid > 0)
>> {
>> ec = args.err;
>> - assert (ec >= 0);
>> if (ec != 0)
>> - __waitpid (new_pid, NULL, 0);
>> + {
>> + /* It handles the unlikely case where the auxiliary process is
>> + terminated by a signal before calling _exit or execve. For
>> + this case the signal is relayed to the called, as the spawn
>> + was successful. */
>> + int status;
>> + __waitpid (new_pid, &status, WNOHANG);
>> + if (WIFSIGNALED (status))
>> + ec = 0;
>> + }
>
> I'm sorry to say, but this does not look correct to me. A SIGCHLD
> handler may have collected the PID at the point of the waitpid call,
> so it can fail with ECHLD or collect the wait notification for another
> subprocess due to PID reuse.
Indeed the it does not handle the pid reuse in the case a SIGCHLD handler
is active. However I think we can't really eliminate all possible races
races in this case (although we can improve its handling).
>
> I think we need to treat ec == -1 (process did not proceed to the
> execve call because it was terminated by a signal) just like ec == 0
> (process made a successful execve call, or was terminated just before
> the system call, or just after a failed execve, again due to signals).
> That is, return the PID to the caller and let it handle the abnormal
> process termination due t a signal.
>
> So the assert needs to be removed and the condition changed from ec !=
> 0 to ec > 0. And this needs plenty of comments, explaining that we
> report subprocess termination due to a signal to the caller and cannot
> handle it locally due to the PID reuse race.
> > I'm not sure if it is necessary to initialize args.err (and ec) to -1
> in the first place and reset it in the subprocess. It seems to me
> that this can be simplified a bit because we only need the non-zero
> value to carry the information that execve failed, and the error code
> with which it failed.
I think the best approach is to just initialize args.err to 0 to indicate
a default successful case and handle the case where the args.err is set to
positive value with WNOHANG. The args.err check sign is not really important,
we just need to actually check for a value different than 0. I think we
will have the following scenarios:
1. For default case with a success execution, args.err will be 0, pid
won't be collected and it will be reported to caller.
2. For default failure case, args.err will be positive and the it will
be collected by the waitpid. An error will be reported to the caller.
3. For the unlikely case where the process was terminated and not
collected by a caller signal handler, it will be reported as a
successful execution and not be collected by posix_spawn (since
args.err will be 0). The caller will need to actually handle this case.
4. For the unlikely case where the process was terminated and collected
by caller we have 3 other possible scenarios:
4.1. The auxiliary process was terminated with args.err equal to 0:
it will handled as 1. (so it does not matter if we hit the pid
reuse race since we won't possible collect an unexpected
process).
4.2. The auxiliary process was terminated after execve (due a failure
in calling it) and before setting args.err to -1: it will also
be handle as 1. but with the issue of not be able to report the
caller a possible execve failures.
4.3. The auxiliary process was terminated after args.err is set to -1:
this is the case where it will be possible to hit the pid reuse
case where we will need to collected the auxiliary pid but we
can not be sure if it will be expected one. I think for this
case we need to actually change waitpid to use WNOHANG to avoid
hanging indefinitely on the call and report an error to caller
since we can't differentiate between a default failure as 2.
and a possible pid reuse race issue.
What about the patch below:
---
As noted by Florian Weimer, current Linux posix_spawn implementation
can trigger an assert if the auxiliary process is terminated before
actually setting the err member:
340 /* Child must set args.err to something non-negative - we rely on
341 the parent and child sharing VM. */
342 args.err = -1;
[...]
362 new_pid = CLONE (__spawni_child, STACK (stack, stack_size), stack_size,
363 CLONE_VM | CLONE_VFORK | SIGCHLD, &args);
364
365 if (new_pid > 0)
366 {
367 ec = args.err;
368 assert (ec >= 0);
Another possible issue is killing the child between setting the err and
actually calling execve. In this case the process will not ran, but
posix_spawn also will not report any error:
269
270 args->err = 0;
271 args->exec (args->file, args->argv, args->envp);
As suggested by Andreas Schwab, this patch removes the faulty assert
and also handles any signal that happens before fork and execve as the
spawn was successful (and thus relaying the handling to the caller to
figure this out). Different than Florian, I can not see why using
atomics to set err would help here, essentially the code runs
sequentially (due CLONE_VFORK) and I think it would not be legal the
compiler evaluate ec without checking for new_pid result (thus there
is no need to compiler barrier).
Summarizing the possible scenarios on posix_spawn execution, we
have:
1. For default case with a success execution, args.err will be 0, pid
will not be collected and it will be reported to caller.
2. For default failure case, args.err will be positive and the it will
be collected by the waitpid. An error will be reported to the
caller.
3. For the unlikely case where the process was terminated and not
collected by a caller signal handler, it will be reported as succeful
execution and not be collected by posix_spawn (since args.err will
be 0). The caller will need to actually handle this case.
4. For the unlikely case where the process was terminated and collected
by caller we have 3 other possible scenarios:
4.1. The auxiliary process was terminated with args.err equal to 0:
it will handled as 1. (so it does not matter if we hit the pid
reuse race since we won't possible collect an unexpected
process).
4.2. The auxiliary process was terminated after execve (due a failure
in calling it) and before setting args.err to -1: it will also
be handle as 1. but with the issue of not be able to report the
caller a possible execve failures.
4.3. The auxiliary process was terminated after args.err is set to -1:
this is the case where it will be possible to hit the pid reuse
case where we will need to collected the auxiliary pid but we
can not be sure if it will be expected one. I think for this
case we need to actually change waitpid to use WNOHANG to avoid
hanging indefinitely on the call and report an error to caller
since we can't differentiate between a default failure as 2.
and a possible pid reuse race issue.
Checked on x86_64-linux-gnu.
* sysdeps/unix/sysv/linux/spawni.c (__spawnix): Handle the case where
the auxiliary process is terminated by a signal before calling _exit
or execve.
---
diff --git a/sysdeps/unix/sysv/linux/spawni.c b/sysdeps/unix/sysv/linux/spawni.c
index dea1650..74c26cb 100644
--- a/sysdeps/unix/sysv/linux/spawni.c
+++ b/sysdeps/unix/sysv/linux/spawni.c
@@ -17,7 +17,6 @@
<http://www.gnu.org/licenses/>. */
#include <spawn.h>
-#include <assert.h>
#include <fcntl.h>
#include <paths.h>
#include <string.h>
@@ -268,7 +267,6 @@ __spawni_child (void *arguments)
__sigprocmask (SIG_SETMASK, (attr->__flags & POSIX_SPAWN_SETSIGMASK)
? &attr->__ss : &args->oldmask, 0);
- args->err = 0;
args->exec (args->file, args->argv, args->envp);
/* This is compatibility function required to enable posix_spawn run
@@ -339,7 +337,7 @@ __spawnix (pid_t * pid, const char *file,
/* Child must set args.err to something non-negative - we rely on
the parent and child sharing VM. */
- args.err = -1;
+ args.err = 0;
args.file = file;
args.exec = exec;
args.fa = file_actions;
@@ -362,12 +360,26 @@ __spawnix (pid_t * pid, const char *file,
new_pid = CLONE (__spawni_child, STACK (stack, stack_size), stack_size,
CLONE_VM | CLONE_VFORK | SIGCHLD, &args);
+ /* It needs to colect the case where the auxiliary process was created
+ but failed to execute the file (due either any preparation step or
+ for execve itself). */
if (new_pid > 0)
{
+ /* Also, it handles the unlikely case where the auxiliary process was
+ terminated before calling execve as it was successfully. The
+ args.err is set to 0 as default and changed to a positive value
+ only in case of failure, so in case of premature termination
+ due a signal args.err will remain zeroed and it will be up to
+ caller to actually collected it. */
ec = args.err;
- assert (ec >= 0);
- if (ec != 0)
- __waitpid (new_pid, NULL, 0);
+ if (ec > 0)
+ /* There still an unlikely case where the child is cancelled after
+ setting args.to a positive error value. Also due a possible
+ pid reuse race (where the kernel allocated the same pid to
+ unrelated process) we need not to undefinitely hang expecting
+ an invalid pid. In both cases an error is returned to the
+ caller. */
+ __waitpid (new_pid, NULL, WNOHANG);
}
else
ec = -new_pid;
--
2.7.4