When we want to duplicate a new process, dup_task_struct() will undergo
a series of allocations. If alloc_thread_info_node() fails, we call
free_task_struct() and return.
This seems right, but it is not. free_task_struct() will not only free
the task struct from the kmem_cache, but will also call
arch_release_task_struct(). The problem is that this function is
supposed to undo whatever arch-specific work done by
arch_dup_task_struct(), that is not yet called at this point. The
particular problem I ran accross was that in x86, we will arrive at
fpu_free() without having ever allocated it.
Signed-off-by: Glauber Costa <[email protected]>
Reported-by: Frederic Weisbecker <[email protected]>
CC: Thomas Gleixner <[email protected]>
CC: Tony Luck <[email protected]>
CC: Fenghua Yu <[email protected]>
---
arch/ia64/include/asm/thread_info.h | 2 +-
kernel/fork.c | 9 +++++++--
2 files changed, 8 insertions(+), 3 deletions(-)
diff --git a/arch/ia64/include/asm/thread_info.h b/arch/ia64/include/asm/thread_info.h
index f7ee853..62452bd 100644
--- a/arch/ia64/include/asm/thread_info.h
+++ b/arch/ia64/include/asm/thread_info.h
@@ -90,7 +90,7 @@ struct thread_info {
\
ret; \
})
-#define free_task_struct(tsk) free_pages((unsigned long) (tsk), KERNEL_STACK_SIZE_ORDER)
+#define __free_task_struct(tsk) free_pages((unsigned long) (tsk), KERNEL_STACK_SIZE_ORDER)
#endif /* !__ASSEMBLY */
diff --git a/kernel/fork.c b/kernel/fork.c
index 152d023..76acb1a 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -124,10 +124,15 @@ static inline struct task_struct *alloc_task_struct_node(int node)
void __weak arch_release_task_struct(struct task_struct *tsk) { }
+static inline void __free_task_struct(struct task_struct *tsk)
+{
+ kmem_cache_free(task_struct_cachep, tsk);
+}
+
static inline void free_task_struct(struct task_struct *tsk)
{
arch_release_task_struct(tsk);
- kmem_cache_free(task_struct_cachep, tsk);
+ __free_task_struct(tsk);
}
#endif
@@ -299,7 +304,7 @@ static struct task_struct *dup_task_struct(struct task_struct *orig)
ti = alloc_thread_info_node(tsk, node);
if (!ti) {
- free_task_struct(tsk);
+ __free_task_struct(tsk);
return NULL;
}
--
1.7.11.4
On Thu, Aug 23, 2012 at 07:36:08PM +0400, Glauber Costa wrote:
> When we want to duplicate a new process, dup_task_struct() will undergo
> a series of allocations. If alloc_thread_info_node() fails, we call
> free_task_struct() and return.
>
> This seems right, but it is not. free_task_struct() will not only free
> the task struct from the kmem_cache, but will also call
> arch_release_task_struct(). The problem is that this function is
> supposed to undo whatever arch-specific work done by
> arch_dup_task_struct(), that is not yet called at this point. The
> particular problem I ran accross was that in x86, we will arrive at
> fpu_free() without having ever allocated it.
>
> Signed-off-by: Glauber Costa <[email protected]>
> Reported-by: Frederic Weisbecker <[email protected]>
Tested-by: Frederic Weisbecker <[email protected]>
On Thu, 23 Aug 2012 19:36:08 +0400
Glauber Costa <[email protected]> wrote:
> When we want to duplicate a new process, dup_task_struct() will undergo
> a series of allocations. If alloc_thread_info_node() fails, we call
> free_task_struct() and return.
>
> This seems right, but it is not. free_task_struct() will not only free
> the task struct from the kmem_cache, but will also call
> arch_release_task_struct(). The problem is that this function is
> supposed to undo whatever arch-specific work done by
> arch_dup_task_struct(), that is not yet called at this point. The
> particular problem I ran accross was that in x86, we will arrive at
> fpu_free() without having ever allocated it.
I think ths was already fixed by f19b9f74b7ea3b ("fork: fix error
handling in dup_task()"). As you would have noticed if you were
preparing patches against up-to-date kernel versions!
On 08/24/2012 06:08 PM, Andrew Morton wrote:
> On Thu, 23 Aug 2012 19:36:08 +0400
> Glauber Costa <[email protected]> wrote:
>
>> When we want to duplicate a new process, dup_task_struct() will undergo
>> a series of allocations. If alloc_thread_info_node() fails, we call
>> free_task_struct() and return.
>>
>> This seems right, but it is not. free_task_struct() will not only free
>> the task struct from the kmem_cache, but will also call
>> arch_release_task_struct(). The problem is that this function is
>> supposed to undo whatever arch-specific work done by
>> arch_dup_task_struct(), that is not yet called at this point. The
>> particular problem I ran accross was that in x86, we will arrive at
>> fpu_free() without having ever allocated it.
>
> I think ths was already fixed by f19b9f74b7ea3b ("fork: fix error
> handling in dup_task()"). As you would have noticed if you were
> preparing patches against up-to-date kernel versions!
>
I am basing all my patches against mmotm (actually, Michal's git copy of
it...)
I might have missed one spin, though. It happens