2023-03-23 09:47:05

by Gou Hao

Subject: [PATCH] fs/buffer: adjust the order of might_sleep() in __getblk_gfp()

From: Gou Hao <[email protected]>

If 'bh' is found in cache, just return directly.
might_sleep() is only required on slow paths.

Signed-off-by: Gou Hao <[email protected]>
---
fs/buffer.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/fs/buffer.c b/fs/buffer.c
index 9e1e2add541e..e13eff504fb5 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -1371,10 +1371,11 @@ __getblk_gfp(struct block_device *bdev, sector_t block,
{
struct buffer_head *bh = __find_get_block(bdev, block, size);

+ if (bh)
+ return bh;
+
might_sleep();
- if (bh == NULL)
- bh = __getblk_slow(bdev, block, size, gfp);
- return bh;
+ return __getblk_slow(bdev, block, size, gfp);
}
EXPORT_SYMBOL(__getblk_gfp);

--
2.20.1

2023-03-23 12:31:57

by Matthew Wilcox

Subject: Re: [PATCH] fs/buffer: adjust the order of might_sleep() in __getblk_gfp()

On Thu, Mar 23, 2023 at 05:37:52PM +0800, [email protected] wrote:
> From: Gou Hao <[email protected]>
>
> If 'bh' is found in cache, just return directly.
> might_sleep() is only required on slow paths.

You're missing the point. The caller can't know whether the slow or
fast path will be taken. So it must _never_ call this function if it
cannot sleep.

2023-03-24 01:55:23

by Gou Hao

Subject: Re: [PATCH] fs/buffer: adjust the order of might_sleep() in __getblk_gfp()

>On Thu, Mar 23, 2023 at 05:37:52PM +0800, [email protected] wrote:
>> From: Gou Hao <[email protected]>
>>
>> If 'bh' is found in cache, just return directly.
>> might_sleep() is only required on slow paths.
>
>You're missing the point. The caller can't know whether the slow or
>fast path will be taken. So it must _never_ call this function if it
>cannot sleep.
I see. Thank you for your explanation! :)