On 2012-07-27 22:21 Jeff Moyer <jmoyer@xxxxxxxxxx> Wrote:
>majianpeng <majianpeng@xxxxxxxxx> writes:
>
>> When exec bio_alloc, the bi_rw is zero.But after calling bio_add_page,
>> it will use bi_rw.
>> Fox example, in functiion __bio_add_page,it will call merge_bvec_fn().
>> The merge_bvec_fn of raid456 will use the bi_rw to judge the merge.
>>>> if ((bvm->bi_rw & 1) == WRITE)
>>>> return biovec->bv_len; /* always allow writes to be mergeable */
>>
>> Signed-off-by: Jianpeng Ma <majianpeng@xxxxxxxxx>
>
>Good catch. How did you find this? Did you experience data corruption
>as a result of this oversight, reduced performance due to missed merge
>opportunities, or did you just notice it in reviewing the code?
>
>Reviewed-by: Jeff Moyer <jmoyer@xxxxxxxxxx>
>
Sorry for late to reply. When i analysed the performance of raid5, i found this bug.
>>
>> There are many place like this in kernel.If you think this patch ok, i will correct those.
>> ---
>> fs/direct-io.c | 1 +
>> 1 files changed, 1 insertions(+), 0 deletions(-)
>>
>> diff --git a/fs/direct-io.c b/fs/direct-io.c
>> index 1faf4cb..77f0bbf 100644
>> --- a/fs/direct-io.c
>> +++ b/fs/direct-io.c
>> @@ -349,6 +349,7 @@ dio_bio_alloc(struct dio *dio, struct dio_submit *sdio,
>>
>> bio->bi_bdev = bdev;
>> bio->bi_sector = first_sector;
>> + bio->bi_rw = dio->rw;
>> if (dio->is_async)
>> bio->bi_end_io = dio_bio_end_aio;
>> else?韬{.n?????%??檩??w?{.n???{饼?z鳐??骅w* jg????????G??⒏ ⒎?:+v???????? ????＂??????
