2008-06-01 16:53:34

by Oleg Nesterov

Subject: [PATCH] coredump: binfmt_elf{,_fdpic}.c: don't abuse GFP_ATOMIC

kmalloc(GFP_ATOMIC) was needed because it was called under tasklist_lock.
The code was changed to use rcu_read_lock(), we can use GFP_KERNEL.

Signed-off-by: Oleg Nesterov <[email protected]>

binfmt_elf.c | 4 ++--
binfmt_elf_fdpic.c | 2 +-
2 files changed, 3 insertions(+), 3 deletions(-)

--- 26-rc2/fs/binfmt_elf.c~NO_ATOMIC 2008-05-18 15:43:59.000000000 +0400
+++ 26-rc2/fs/binfmt_elf.c 2008-06-01 20:47:08.000000000 +0400
@@ -1524,7 +1524,7 @@ static int fill_note_info(struct elfhdr
if (p->mm == dump_task->mm) {
t = kzalloc(offsetof(struct elf_thread_core_info,
notes[info->thread_notes]),
- GFP_ATOMIC);
+ GFP_KERNEL);
if (unlikely(!t)) {
rcu_read_unlock();
return 0;
@@ -1726,7 +1726,7 @@ static int fill_note_info(struct elfhdr
rcu_read_lock();
do_each_thread(g, p)
if (current->mm == p->mm && current != p) {
- ets = kzalloc(sizeof(*ets), GFP_ATOMIC);
+ ets = kzalloc(sizeof(*ets), GFP_KERNEL);
if (!ets) {
rcu_read_unlock();
return 0;
--- 26-rc2/fs/binfmt_elf_fdpic.c~NO_ATOMIC 2008-05-18 15:43:59.000000000 +0400
+++ 26-rc2/fs/binfmt_elf_fdpic.c 2008-06-01 20:47:20.000000000 +0400
@@ -1626,7 +1626,7 @@ static int elf_fdpic_core_dump(long sign
rcu_read_lock();
do_each_thread(g,p)
if (current->mm == p->mm && current != p) {
- tmp = kzalloc(sizeof(*tmp), GFP_ATOMIC);
+ tmp = kzalloc(sizeof(*tmp), GFP_KERNEL);
if (!tmp) {
rcu_read_unlock();
goto cleanup;

2008-06-01 20:23:47

by Peter Zijlstra

Subject: Re: [PATCH] coredump: binfmt_elf{,_fdpic}.c: don't abuse GFP_ATOMIC

On Sun, 2008-06-01 at 20:55 +0400, Oleg Nesterov wrote:
> kmalloc(GFP_ATOMIC) was needed because it was called under tasklist_lock.
> The code was changed to use rcu_read_lock(), we can use GFP_KERNEL.

Classic RCU - still frequently used - disables preemption and thus does
not allow us to use __GFP_WAIT.

> Signed-off-by: Oleg Nesterov <[email protected]>
>
> binfmt_elf.c | 4 ++--
> binfmt_elf_fdpic.c | 2 +-
> 2 files changed, 3 insertions(+), 3 deletions(-)
>
> --- 26-rc2/fs/binfmt_elf.c~NO_ATOMIC 2008-05-18 15:43:59.000000000 +0400
> +++ 26-rc2/fs/binfmt_elf.c 2008-06-01 20:47:08.000000000 +0400
> @@ -1524,7 +1524,7 @@ static int fill_note_info(struct elfhdr
> if (p->mm == dump_task->mm) {
> t = kzalloc(offsetof(struct elf_thread_core_info,
> notes[info->thread_notes]),
> - GFP_ATOMIC);
> + GFP_KERNEL);
> if (unlikely(!t)) {
> rcu_read_unlock();
> return 0;
> @@ -1726,7 +1726,7 @@ static int fill_note_info(struct elfhdr
> rcu_read_lock();
> do_each_thread(g, p)
> if (current->mm == p->mm && current != p) {
> - ets = kzalloc(sizeof(*ets), GFP_ATOMIC);
> + ets = kzalloc(sizeof(*ets), GFP_KERNEL);
> if (!ets) {
> rcu_read_unlock();
> return 0;
> --- 26-rc2/fs/binfmt_elf_fdpic.c~NO_ATOMIC 2008-05-18 15:43:59.000000000 +0400
> +++ 26-rc2/fs/binfmt_elf_fdpic.c 2008-06-01 20:47:20.000000000 +0400
> @@ -1626,7 +1626,7 @@ static int elf_fdpic_core_dump(long sign
> rcu_read_lock();
> do_each_thread(g,p)
> if (current->mm == p->mm && current != p) {
> - tmp = kzalloc(sizeof(*tmp), GFP_ATOMIC);
> + tmp = kzalloc(sizeof(*tmp), GFP_KERNEL);
> if (!tmp) {
> rcu_read_unlock();
> goto cleanup;
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/

2008-06-01 21:09:03

by Oleg Nesterov

Subject: Re: [PATCH] coredump: binfmt_elf{,_fdpic}.c: don't abuse GFP_ATOMIC

On 06/01, Peter Zijlstra wrote:
>
> On Sun, 2008-06-01 at 20:55 +0400, Oleg Nesterov wrote:
> > kmalloc(GFP_ATOMIC) was needed because it was called under tasklist_lock.
> > The code was changed to use rcu_read_lock(), we can use GFP_KERNEL.
>
> Classic RCU - still frequently used - disables preemption and thus does
> not allow us to use __GFP_WAIT.

Ah, I'm stupid, thanks!

please ignore this patch.

Oleg.