On 5/3/18 11:36 AM, Jeff Moyer wrote: > Hi, Adam, Hello Jeff, > > adam.manzanares@xxxxxxx writes: > >> From: Adam Manzanares <adam.manzanares@xxxxxxx> >> >> This is the per-I/O equivalent of the ioprio_set system call. >> >> When IOCB_FLAG_IOPRIO is set on the iocb aio_flags field, then we set the >> newly added kiocb ki_ioprio field to the value in the iocb aio_reqprio field. >> >> When a bio is created for an aio request by the block dev we set the priority >> value of the bio to the user supplied value. >> >> See the following link for performance implications on a SATA HDD: >> https://lkml.org/lkml/2016/12/6/495 >> >> Given that WRR support for NVME devices has patches floating around and it was >> discussed at LSFMM, we may soon have a lower latency storage device that can >> make use of iopriorities. A per command iopriority interface seems timely >> given these developments. >> >> If we want to avoid bloating struct kiocb, I suggest we turn the private field >> into a union of the private and ki_ioprio field. It seems like the users of >> the private field all use it at a point where we can yank the priority from >> the kiocb before the private field is used. Comments and suggestions welcome. > > The ioprio_set system call requires CAP_SYS_ADMIN for setting > IOPRIO_CLASS_RT. I think we need similar checks here. I forgot how dangerous IOPRIO_CLASS_RT can be :). I will make a new function based on the checks in ioprio.c and reuse. Thanks, Adam > > -Jeff > >> >> v2: merge patches >> use IOCB_FLAG_IOPRIO >> validate intended use with IOCB_IOPRIO >> add linux-api and linux-block to cc >> >> Signed-off-by: Adam Manzanares <adam.manzanares@xxxxxxx> >> --- >> fs/aio.c | 10 ++++++++++ >> fs/block_dev.c | 2 ++ >> include/linux/fs.h | 2 ++ >> include/uapi/linux/aio_abi.h | 1 + >> 4 files changed, 15 insertions(+) >> >> diff --git a/fs/aio.c b/fs/aio.c >> index 88d7927ffbc6..f36636d8ff2c 100644 >> --- a/fs/aio.c >> +++ b/fs/aio.c >> @@ -1597,6 +1597,16 @@ static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb, >> req->common.ki_flags |= IOCB_EVENTFD; >> } >> >> + if (iocb->aio_flags & IOCB_FLAG_IOPRIO) { >> + /* >> + * If the IOCB_FLAG_IOPRIO flag of aio_flags is set, then >> + * aio_reqprio is interpreted as an I/O scheduling >> + * class and priority. >> + */ >> + req->common.ki_ioprio = iocb->aio_reqprio; >> + req->common.ki_flags |= IOCB_IOPRIO; >> + } >> + >> ret = kiocb_set_rw_flags(&req->common, iocb->aio_rw_flags); >> if (unlikely(ret)) { >> pr_debug("EINVAL: aio_rw_flags\n"); >> diff --git a/fs/block_dev.c b/fs/block_dev.c >> index 7ec920e27065..970bef79caa6 100644 >> --- a/fs/block_dev.c >> +++ b/fs/block_dev.c >> @@ -355,6 +355,8 @@ __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter, int nr_pages) >> bio->bi_write_hint = iocb->ki_hint; >> bio->bi_private = dio; >> bio->bi_end_io = blkdev_bio_end_io; >> + if (iocb->ki_flags & IOCB_IOPRIO) >> + bio->bi_ioprio = iocb->ki_ioprio; >> >> ret = bio_iov_iter_get_pages(bio, iter); >> if (unlikely(ret)) { >> diff --git a/include/linux/fs.h b/include/linux/fs.h >> index 760d8da1b6c7..ab63ce720305 100644 >> --- a/include/linux/fs.h >> +++ b/include/linux/fs.h >> @@ -292,6 +292,7 @@ enum rw_hint { >> #define IOCB_SYNC (1 << 5) >> #define IOCB_WRITE (1 << 6) >> #define IOCB_NOWAIT (1 << 7) >> +#define IOCB_IOPRIO (1 << 8) >> >> struct kiocb { >> struct file *ki_filp; >> @@ -300,6 +301,7 @@ struct kiocb { >> void *private; >> int ki_flags; >> enum rw_hint ki_hint; >> + u16 ki_ioprio; /* See linux/ioprio.h */ >> } __randomize_layout; >> >> static inline bool is_sync_kiocb(struct kiocb *kiocb) >> diff --git a/include/uapi/linux/aio_abi.h b/include/uapi/linux/aio_abi.h >> index a04adbc70ddf..d4593a6062ef 100644 >> --- a/include/uapi/linux/aio_abi.h >> +++ b/include/uapi/linux/aio_abi.h >> @@ -54,6 +54,7 @@ enum { >> * is valid. >> */ >> #define IOCB_FLAG_RESFD (1 << 0) >> +#define IOCB_FLAG_IOPRIO (1 << 1) >> >> /* read() from /dev/aio returns these structures. */ >> struct io_event {��.n��������+%������w��{.n�����{����* jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
