Commit 8deb24dcb89cb ("pid: make ENOMEM return value more obvious")
left the return value uninitialized in one error case. The justification
for the above commit included a statement that retval is "initialized on
ever[y] failure path in the loop". However, that is not quite good
enough because there is an earlier case that is before the loop. And
also, it's more maintenance and merge-safe to initialize it once at the
top, as evidenced by this build warning that we now have.
Therefore, restore the top-level initialization of retval.
Also move the descriptive comment up, and remove the now-redundant
later initialization of retval.
Fixes: 8deb24dcb89cb ("pid: make ENOMEM return value more obvious")
Cc: Christian Brauner <[email protected]>
Signed-off-by: John Hubbard <[email protected]>
---
kernel/pid.c | 21 ++++++++++-----------
1 file changed, 10 insertions(+), 11 deletions(-)
diff --git a/kernel/pid.c b/kernel/pid.c
index ff6cd6786d10..49bf6dd32de4 100644
--- a/kernel/pid.c
+++ b/kernel/pid.c
@@ -162,7 +162,16 @@ struct pid *alloc_pid(struct pid_namespace *ns, pid_t *set_tid,
int i, nr;
struct pid_namespace *tmp;
struct upid *upid;
- int retval;
+
+ /*
+ * ENOMEM is not the most obvious choice especially for the case
+ * where the child subreaper has already exited and the pid
+ * namespace denies the creation of any new processes. But ENOMEM
+ * is what we have exposed to userspace for a long time and it is
+ * documented behavior for pid namespaces. So we can't easily
+ * change it even if there were an error code better suited.
+ */
+ int retval = -ENOMEM;
/*
* set_tid_size contains the size of the set_tid array. Starting at
@@ -244,16 +253,6 @@ struct pid *alloc_pid(struct pid_namespace *ns, pid_t *set_tid,
tmp = tmp->parent;
}
- /*
- * ENOMEM is not the most obvious choice especially for the case
- * where the child subreaper has already exited and the pid
- * namespace denies the creation of any new processes. But ENOMEM
- * is what we have exposed to userspace for a long time and it is
- * documented behavior for pid namespaces. So we can't easily
- * change it even if there were an error code better suited.
- */
- retval = -ENOMEM;
-
get_pid_ns(ns);
refcount_set(&pid->count, 1);
for (type = 0; type < PIDTYPE_MAX; ++type)
base-commit: 134546626849cd6852d6d4bf8f207b5fbc54261b
--
2.25.1
On Tue, Mar 10, 2020 at 08:21:53PM -0700, John Hubbard wrote:
> Commit 8deb24dcb89cb ("pid: make ENOMEM return value more obvious")
> left the return value uninitialized in one error case. The justification
> for the above commit included a statement that retval is "initialized on
> ever[y] failure path in the loop". However, that is not quite good
> enough because there is an earlier case that is before the loop. And
> also, it's more maintenance and merge-safe to initialize it once at the
> top, as evidenced by this build warning that we now have.
>
> Therefore, restore the top-level initialization of retval.
>
> Also move the descriptive comment up, and remove the now-redundant
> later initialization of retval.
>
> Fixes: 8deb24dcb89cb ("pid: make ENOMEM return value more obvious")
> Cc: Christian Brauner <[email protected]>
> Signed-off-by: John Hubbard <[email protected]>
Thanks. I already have a fixed-up version in my tree after Stephen
reported it earlier today:
https://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux.git/commit/?h=fixes&id=10dab84caf400f2f5f8b010ebb0c7c4272ec5093
(Fwiw, we can't just move the it back up. It needs to be set after the
loop too because it can be set to EPERM before. See
https://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux.git/commit/?h=fixes&id=b26ebfe12f34f372cf041c6f801fa49c3fb382c5
for the motiviation for the comment.)
Christian
On 3/10/20 8:30 PM, Christian Brauner wrote:
> On Tue, Mar 10, 2020 at 08:21:53PM -0700, John Hubbard wrote:
>> Commit 8deb24dcb89cb ("pid: make ENOMEM return value more obvious")
>> left the return value uninitialized in one error case. The justification
>> for the above commit included a statement that retval is "initialized on
>> ever[y] failure path in the loop". However, that is not quite good
>> enough because there is an earlier case that is before the loop. And
>> also, it's more maintenance and merge-safe to initialize it once at the
>> top, as evidenced by this build warning that we now have.
>>
>> Therefore, restore the top-level initialization of retval.
>>
>> Also move the descriptive comment up, and remove the now-redundant
>> later initialization of retval.
>>
>> Fixes: 8deb24dcb89cb ("pid: make ENOMEM return value more obvious")
>> Cc: Christian Brauner <[email protected]>
>> Signed-off-by: John Hubbard <[email protected]>
>
> Thanks. I already have a fixed-up version in my tree after Stephen
> reported it earlier today:
> https://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux.git/commit/?h=fixes&id=10dab84caf400f2f5f8b010ebb0c7c4272ec5093
>
> (Fwiw, we can't just move the it back up. It needs to be set after the
> loop too because it can be set to EPERM before. See
OK, yes. Sounds good.
thanks,
--
John Hubbard
NVIDIA
> https://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux.git/commit/?h=fixes&id=b26ebfe12f34f372cf041c6f801fa49c3fb382c5
> for the motiviation for the comment.)
>
> Christian
>