| Message ID | 20220516162118.155763-4-axboe@kernel.dk (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | Add support for ring mapped provided buffers | expand |
Hi All, On 5/17/22 00:21, Jens Axboe wrote: > Provided buffers allow an application to supply io_uring with buffers > that can then be grabbed for a read/receive request, when the data > source is ready to deliver data. The existing scheme relies on using > IORING_OP_PROVIDE_BUFFERS to do that, but it can be difficult to use > in real world applications. It's pretty efficient if the application > is able to supply back batches of provided buffers when they have been > consumed and the application is ready to recycle them, but if > fragmentation occurs in the buffer space, it can become difficult to > supply enough buffers at the time. This hurts efficiency. > > Add a register op, IORING_REGISTER_PBUF_RING, which allows an application > to setup a shared queue for each buffer group of provided buffers. The > application can then supply buffers simply by adding them to this ring, > and the kernel can consume then just as easily. The ring shares the head > with the application, the tail remains private in the kernel. > > Provided buffers setup with IORING_REGISTER_PBUF_RING cannot use > IORING_OP_{PROVIDE,REMOVE}_BUFFERS for adding or removing entries to the > ring, they must use the mapped ring. Mapped provided buffer rings can > co-exist with normal provided buffers, just not within the same group ID. > > To gauge overhead of the existing scheme and evaluate the mapped ring > approach, a simple NOP benchmark was written. It uses a ring of 128 > entries, and submits/completes 32 at the time. 'Replenish' is how > many buffers are provided back at the time after they have been > consumed: > > Test Replenish NOPs/sec > ================================================================ > No provided buffers NA ~30M > Provided buffers 32 ~16M > Provided buffers 1 ~10M > Ring buffers 32 ~27M > Ring buffers 1 ~27M > > The ring mapped buffers perform almost as well as not using provided > buffers at all, and they don't care if you provided 1 or more back at > the same time. This means application can just replenish as they go, > rather than need to batch and compact, further reducing overhead in the > application. The NOP benchmark above doesn't need to do any compaction, > so that overhead isn't even reflected in the above test. > > Co-developed-by: Dylan Yudaken <dylany@fb.com> > Signed-off-by: Jens Axboe <axboe@kernel.dk> > --- > fs/io_uring.c | 233 ++++++++++++++++++++++++++++++++-- > include/uapi/linux/io_uring.h | 36 ++++++ > 2 files changed, 257 insertions(+), 12 deletions(-) > > diff --git a/fs/io_uring.c b/fs/io_uring.c > index 5867dcabc73b..776a9f5e5ec7 100644 > --- a/fs/io_uring.c > +++ b/fs/io_uring.c > @@ -285,9 +285,26 @@ struct io_rsrc_data { > bool quiesce; > }; > > +#define IO_BUFFER_LIST_BUF_PER_PAGE (PAGE_SIZE / sizeof(struct io_uring_buf)) > struct io_buffer_list { > - struct list_head buf_list; > + /* > + * If ->buf_nr_pages is set, then buf_pages/buf_ring are used. If not, > + * then these are classic provided buffers and ->buf_list is used. > + */ > + union { > + struct list_head buf_list; > + struct { > + struct page **buf_pages; > + struct io_uring_buf_ring *buf_ring; > + }; > + }; > __u16 bgid; > + > + /* below is for ring provided buffers */ > + __u16 buf_nr_pages; > + __u16 nr_entries; > + __u32 tail; > + __u32 mask; > }; > > struct io_buffer { > @@ -804,6 +821,7 @@ enum { > REQ_F_NEED_CLEANUP_BIT, > REQ_F_POLLED_BIT, > REQ_F_BUFFER_SELECTED_BIT, > + REQ_F_BUFFER_RING_BIT, > REQ_F_COMPLETE_INLINE_BIT, > REQ_F_REISSUE_BIT, > REQ_F_CREDS_BIT, > @@ -855,6 +873,8 @@ enum { > REQ_F_POLLED = BIT(REQ_F_POLLED_BIT), > /* buffer already selected */ > REQ_F_BUFFER_SELECTED = BIT(REQ_F_BUFFER_SELECTED_BIT), > + /* buffer selected from ring, needs commit */ > + REQ_F_BUFFER_RING = BIT(REQ_F_BUFFER_RING_BIT), > /* completion is deferred through io_comp_state */ > REQ_F_COMPLETE_INLINE = BIT(REQ_F_COMPLETE_INLINE_BIT), > /* caller should reissue async */ > @@ -979,6 +999,12 @@ struct io_kiocb { > > /* stores selected buf, valid IFF REQ_F_BUFFER_SELECTED is set */ > struct io_buffer *kbuf; > + > + /* > + * stores buffer ID for ring provided buffers, valid IFF > + * REQ_F_BUFFER_RING is set. > + */ > + struct io_buffer_list *buf_list; > }; > > union { > @@ -1470,8 +1496,14 @@ static inline void io_req_set_rsrc_node(struct io_kiocb *req, > > static unsigned int __io_put_kbuf(struct io_kiocb *req, struct list_head *list) > { > - req->flags &= ~REQ_F_BUFFER_SELECTED; > - list_add(&req->kbuf->list, list); > + if (req->flags & REQ_F_BUFFER_RING) { > + if (req->buf_list) > + req->buf_list->tail++; This confused me for some time..seems [tail, head) is the registered bufs that kernel space can leverage? similar to what pipe logic does. how about swaping the name of head and tail, this way setting the kernel as a consumer. But this is just my personal preference.. > + req->flags &= ~REQ_F_BUFFER_RING; > + } else { > + list_add(&req->kbuf->list, list); > + req->flags &= ~REQ_F_BUFFER_SELECTED; > + } > > return IORING_CQE_F_BUFFER | (req->buf_index << IORING_CQE_BUFFER_SHIFT); > } > @@ -1480,7 +1512,7 @@ static inline unsigned int io_put_kbuf_comp(struct io_kiocb *req) > { > lockdep_assert_held(&req->ctx->completion_lock); > > - if (likely(!(req->flags & REQ_F_BUFFER_SELECTED))) > + if (!(req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING))) > return 0; > return __io_put_kbuf(req, &req->ctx->io_buffers_comp); > } > @@ -1490,7 +1522,7 @@ static inline unsigned int io_put_kbuf(struct io_kiocb *req, > { > unsigned int cflags; > > - if (likely(!(req->flags & REQ_F_BUFFER_SELECTED))) > + if (!(req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING))) > return 0; > > /* > @@ -1505,7 +1537,10 @@ static inline unsigned int io_put_kbuf(struct io_kiocb *req, > * We migrate buffers from the comp_list to the issue cache list > * when we need one. > */ > - if (issue_flags & IO_URING_F_UNLOCKED) { > + if (req->flags & REQ_F_BUFFER_RING) { > + /* no buffers to recycle for this case */ > + cflags = __io_put_kbuf(req, NULL); > + } else if (issue_flags & IO_URING_F_UNLOCKED) { > struct io_ring_ctx *ctx = req->ctx; > > spin_lock(&ctx->completion_lock); > @@ -1535,11 +1570,23 @@ static void io_kbuf_recycle(struct io_kiocb *req, unsigned issue_flags) > struct io_buffer_list *bl; > struct io_buffer *buf; > > - if (likely(!(req->flags & REQ_F_BUFFER_SELECTED))) > + if (!(req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING))) > return; > /* don't recycle if we already did IO to this buffer */ > if (req->flags & REQ_F_PARTIAL_IO) > return; > + /* > + * We don't need to recycle for REQ_F_BUFFER_RING, we can just clear > + * the flag and hence ensure that bl->tail doesn't get incremented. > + * If the tail has already been incremented, hang on to it. > + */ > + if (req->flags & REQ_F_BUFFER_RING) { > + if (req->buf_list) { > + req->buf_index = req->buf_list->bgid; > + req->flags &= ~REQ_F_BUFFER_RING; > + } > + return; > + } > > io_ring_submit_lock(ctx, issue_flags); > > @@ -3487,6 +3534,53 @@ static void __user *io_provided_buffer_select(struct io_kiocb *req, size_t *len, > return ret; > } > > +static void __user *io_ring_buffer_select(struct io_kiocb *req, size_t *len, > + struct io_buffer_list *bl, > + unsigned int issue_flags) > +{ > + struct io_uring_buf_ring *br = bl->buf_ring; > + struct io_uring_buf *buf; > + __u32 tail = bl->tail; > + > + if (unlikely(smp_load_acquire(&br->head) == tail)) { > + io_ring_submit_unlock(req->ctx, issue_flags); > + return ERR_PTR(-ENOBUFS); > + } > + > + tail &= bl->mask; > + if (tail < IO_BUFFER_LIST_BUF_PER_PAGE) { > + buf = &br->bufs[tail]; > + } else { > + int off = tail & (IO_BUFFER_LIST_BUF_PER_PAGE - 1); > + int index = tail / IO_BUFFER_LIST_BUF_PER_PAGE - 1; Could we do some bitwise trick with some compiler check there since for now IO_BUFFER_LIST_BUF_PER_PAGE is a power of 2. > + buf = page_address(bl->buf_pages[index]); > + buf += off; > + } I'm not familiar with this part, allow me to ask, is this if else statement for efficiency? why choose one page as the dividing line Regards, Hao > + if (*len > buf->len) > + *len = buf->len; > + req->flags |= REQ_F_BUFFER_RING; > + req->buf_list = bl; > + req->buf_index = buf->bid; > + > + if (!(issue_flags & IO_URING_F_UNLOCKED)) > + return u64_to_user_ptr(buf->addr); > + > + /* > + * If we came in unlocked, we have no choice but to > + * consume the buffer here. This does mean it'll be > + * pinned until the IO completes. But coming in > + * unlocked means we're in io-wq context, hence there > + * should be no further retry. For the locked case, the > + * caller must ensure to call the commit when the > + * transfer completes (or if we get -EAGAIN and must > + * poll or retry). > + */ > + req->buf_list = NULL; > + bl->tail++; > + io_ring_submit_unlock(req->ctx, issue_flags); > + return u64_to_user_ptr(buf->addr); > +} > + > static void __user *io_buffer_select(struct io_kiocb *req, size_t *len, > unsigned int issue_flags) > { > @@ -3502,6 +3596,9 @@ static void __user *io_buffer_select(struct io_kiocb *req, size_t *len, > } > > /* selection helpers drop the submit lock again, if needed */ > + if (bl->buf_nr_pages) > + return io_ring_buffer_select(req, len, bl, issue_flags); > + > return io_provided_buffer_select(req, len, bl, issue_flags); > } > > @@ -3558,7 +3655,7 @@ static ssize_t __io_iov_buffer_select(struct io_kiocb *req, struct iovec *iov, > static ssize_t io_iov_buffer_select(struct io_kiocb *req, struct iovec *iov, > unsigned int issue_flags) > { > - if (req->flags & REQ_F_BUFFER_SELECTED) { > + if (req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING)) { > iov[0].iov_base = u64_to_user_ptr(req->rw.addr); > iov[0].iov_len = req->rw.len; > return 0; > @@ -3578,7 +3675,7 @@ static inline bool io_do_buffer_select(struct io_kiocb *req) > { > if (!(req->flags & REQ_F_BUFFER_SELECT)) > return false; > - return !(req->flags & REQ_F_BUFFER_SELECTED); > + return !(req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING)); > } > > static struct iovec *__io_import_iovec(int rw, struct io_kiocb *req, > @@ -4872,6 +4969,17 @@ static int __io_remove_buffers(struct io_ring_ctx *ctx, > if (!nbufs) > return 0; > > + if (bl->buf_nr_pages) { > + int j; > + > + for (j = 0; j < bl->buf_nr_pages; j++) > + unpin_user_page(bl->buf_pages[j]); > + kvfree(bl->buf_pages); > + bl->buf_pages = NULL; > + bl->buf_nr_pages = 0; > + return bl->buf_ring->head - bl->tail; > + } > + > /* the head kbuf is the list itself */ > while (!list_empty(&bl->buf_list)) { > struct io_buffer *nxt; > @@ -4898,8 +5006,12 @@ static int io_remove_buffers(struct io_kiocb *req, unsigned int issue_flags) > > ret = -ENOENT; > bl = io_buffer_get_list(ctx, p->bgid); > - if (bl) > - ret = __io_remove_buffers(ctx, bl, p->nbufs); > + if (bl) { > + ret = -EINVAL; > + /* can't use provide/remove buffers command on mapped buffers */ > + if (!bl->buf_nr_pages) > + ret = __io_remove_buffers(ctx, bl, p->nbufs); > + } > if (ret < 0) > req_set_fail(req); > > @@ -5047,7 +5159,7 @@ static int io_provide_buffers(struct io_kiocb *req, unsigned int issue_flags) > > bl = io_buffer_get_list(ctx, p->bgid); > if (unlikely(!bl)) { > - bl = kmalloc(sizeof(*bl), GFP_KERNEL); > + bl = kzalloc(sizeof(*bl), GFP_KERNEL); > if (!bl) { > ret = -ENOMEM; > goto err; > @@ -5058,6 +5170,11 @@ static int io_provide_buffers(struct io_kiocb *req, unsigned int issue_flags) > goto err; > } > } > + /* can't add buffers via this command for a mapped buffer ring */ > + if (bl->buf_nr_pages) { > + ret = -EINVAL; > + goto err; > + } > > ret = io_add_buffers(ctx, p, bl); > err: > @@ -12011,6 +12128,83 @@ static __cold int io_register_iowq_max_workers(struct io_ring_ctx *ctx, > return ret; > } > > +static int io_register_pbuf_ring(struct io_ring_ctx *ctx, void __user *arg) > +{ > + struct io_uring_buf_ring *br; > + struct io_uring_buf_reg reg; > + struct io_buffer_list *bl; > + struct page **pages; > + int nr_pages; > + > + if (copy_from_user(®, arg, sizeof(reg))) > + return -EFAULT; > + > + if (reg.pad || reg.resv[0] || reg.resv[1] || reg.resv[2]) > + return -EINVAL; > + if (!reg.ring_addr) > + return -EFAULT; > + if (reg.ring_addr & ~PAGE_MASK) > + return -EINVAL; > + if (!is_power_of_2(reg.ring_entries)) > + return -EINVAL; > + > + if (unlikely(reg.bgid < BGID_ARRAY && !ctx->io_bl)) { > + int ret = io_init_bl_list(ctx); > + if (ret) > + return ret; > + } > + > + bl = io_buffer_get_list(ctx, reg.bgid); > + if (bl && bl->buf_nr_pages) > + return -EEXIST; > + if (!bl) { > + bl = kzalloc(sizeof(*bl), GFP_KERNEL); > + if (!bl) > + return -ENOMEM; > + } > + > + pages = io_pin_pages(reg.ring_addr, > + struct_size(br, bufs, reg.ring_entries), > + &nr_pages); > + if (IS_ERR(pages)) { > + kfree(bl); > + return PTR_ERR(pages); > + } > + > + br = page_address(pages[0]); > + bl->buf_pages = pages; > + bl->buf_nr_pages = nr_pages; > + bl->nr_entries = reg.ring_entries; > + bl->buf_ring = br; > + bl->mask = reg.ring_entries - 1; > + io_buffer_add_list(ctx, bl, reg.bgid); > + return 0; > +} > + > +static int io_unregister_pbuf_ring(struct io_ring_ctx *ctx, void __user *arg) > +{ > + struct io_uring_buf_reg reg; > + struct io_buffer_list *bl; > + > + if (copy_from_user(®, arg, sizeof(reg))) > + return -EFAULT; > + if (reg.pad || reg.resv[0] || reg.resv[1] || reg.resv[2]) > + return -EINVAL; > + > + bl = io_buffer_get_list(ctx, reg.bgid); > + if (!bl) > + return -ENOENT; > + if (!bl->buf_nr_pages) > + return -EINVAL; > + > + __io_remove_buffers(ctx, bl, -1U); > + if (bl->bgid >= BGID_ARRAY) { > + xa_erase(&ctx->io_bl_xa, bl->bgid); > + kfree(bl); > + } > + return 0; > +} > + > static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode, > void __user *arg, unsigned nr_args) > __releases(ctx->uring_lock) > @@ -12142,6 +12336,18 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode, > case IORING_UNREGISTER_RING_FDS: > ret = io_ringfd_unregister(ctx, arg, nr_args); > break; > + case IORING_REGISTER_PBUF_RING: > + ret = -EINVAL; > + if (!arg || nr_args != 1) > + break; > + ret = io_register_pbuf_ring(ctx, arg); > + break; > + case IORING_UNREGISTER_PBUF_RING: > + ret = -EINVAL; > + if (!arg || nr_args != 1) > + break; > + ret = io_unregister_pbuf_ring(ctx, arg); > + break; > default: > ret = -EINVAL; > break; > @@ -12227,6 +12433,9 @@ static int __init io_uring_init(void) > /* ->buf_index is u16 */ > BUILD_BUG_ON(IORING_MAX_REG_BUFFERS >= (1u << 16)); > BUILD_BUG_ON(BGID_ARRAY * sizeof(struct io_buffer_list) > PAGE_SIZE); > + BUILD_BUG_ON(offsetof(struct io_uring_buf_ring, bufs) != 0); > + BUILD_BUG_ON(offsetof(struct io_uring_buf, resv) != > + offsetof(struct io_uring_buf_ring, head)); > > /* should fit into one byte */ > BUILD_BUG_ON(SQE_VALID_FLAGS >= (1 << 8)); > diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h > index 15f821af9242..90d78428317a 100644 > --- a/include/uapi/linux/io_uring.h > +++ b/include/uapi/linux/io_uring.h > @@ -384,6 +384,10 @@ enum { > IORING_REGISTER_RING_FDS = 20, > IORING_UNREGISTER_RING_FDS = 21, > > + /* register ring based provide buffer group */ > + IORING_REGISTER_PBUF_RING = 22, > + IORING_UNREGISTER_PBUF_RING = 23, > + > /* this goes last */ > IORING_REGISTER_LAST > }; > @@ -461,6 +465,38 @@ struct io_uring_restriction { > __u32 resv2[3]; > }; > > +struct io_uring_buf { > + __u64 addr; > + __u32 len; > + __u16 bid; > + __u16 resv; > +}; > + > +struct io_uring_buf_ring { > + union { > + /* > + * To avoid spilling into more pages than we need to, the > + * ring head is overlaid with the io_uring_buf->resv field. > + */ > + struct { > + __u64 resv1; > + __u32 resv2; > + __u16 resv3; > + __u16 head; > + }; > + struct io_uring_buf bufs[0]; > + }; > +}; > + > +/* argument for IORING_(UN)REGISTER_PBUF_RING */ > +struct io_uring_buf_reg { > + __u64 ring_addr; > + __u32 ring_entries; > + __u16 bgid; > + __u16 pad; > + __u64 resv[3]; > +}; > + > /* > * io_uring_restriction->opcode values > */
On 5/17/22 8:18 AM, Hao Xu wrote: > Hi All, > > On 5/17/22 00:21, Jens Axboe wrote: >> Provided buffers allow an application to supply io_uring with buffers >> that can then be grabbed for a read/receive request, when the data >> source is ready to deliver data. The existing scheme relies on using >> IORING_OP_PROVIDE_BUFFERS to do that, but it can be difficult to use >> in real world applications. It's pretty efficient if the application >> is able to supply back batches of provided buffers when they have been >> consumed and the application is ready to recycle them, but if >> fragmentation occurs in the buffer space, it can become difficult to >> supply enough buffers at the time. This hurts efficiency. >> >> Add a register op, IORING_REGISTER_PBUF_RING, which allows an application >> to setup a shared queue for each buffer group of provided buffers. The >> application can then supply buffers simply by adding them to this ring, >> and the kernel can consume then just as easily. The ring shares the head >> with the application, the tail remains private in the kernel. >> >> Provided buffers setup with IORING_REGISTER_PBUF_RING cannot use >> IORING_OP_{PROVIDE,REMOVE}_BUFFERS for adding or removing entries to the >> ring, they must use the mapped ring. Mapped provided buffer rings can >> co-exist with normal provided buffers, just not within the same group ID. >> >> To gauge overhead of the existing scheme and evaluate the mapped ring >> approach, a simple NOP benchmark was written. It uses a ring of 128 >> entries, and submits/completes 32 at the time. 'Replenish' is how >> many buffers are provided back at the time after they have been >> consumed: >> >> Test Replenish NOPs/sec >> ================================================================ >> No provided buffers NA ~30M >> Provided buffers 32 ~16M >> Provided buffers 1 ~10M >> Ring buffers 32 ~27M >> Ring buffers 1 ~27M >> >> The ring mapped buffers perform almost as well as not using provided >> buffers at all, and they don't care if you provided 1 or more back at >> the same time. This means application can just replenish as they go, >> rather than need to batch and compact, further reducing overhead in the >> application. The NOP benchmark above doesn't need to do any compaction, >> so that overhead isn't even reflected in the above test. >> >> Co-developed-by: Dylan Yudaken <dylany@fb.com> >> Signed-off-by: Jens Axboe <axboe@kernel.dk> >> --- >> fs/io_uring.c | 233 ++++++++++++++++++++++++++++++++-- >> include/uapi/linux/io_uring.h | 36 ++++++ >> 2 files changed, 257 insertions(+), 12 deletions(-) >> >> diff --git a/fs/io_uring.c b/fs/io_uring.c >> index 5867dcabc73b..776a9f5e5ec7 100644 >> --- a/fs/io_uring.c >> +++ b/fs/io_uring.c >> @@ -285,9 +285,26 @@ struct io_rsrc_data { >> bool quiesce; >> }; >> +#define IO_BUFFER_LIST_BUF_PER_PAGE (PAGE_SIZE / sizeof(struct io_uring_buf)) >> struct io_buffer_list { >> - struct list_head buf_list; >> + /* >> + * If ->buf_nr_pages is set, then buf_pages/buf_ring are used. If not, >> + * then these are classic provided buffers and ->buf_list is used. >> + */ >> + union { >> + struct list_head buf_list; >> + struct { >> + struct page **buf_pages; >> + struct io_uring_buf_ring *buf_ring; >> + }; >> + }; >> __u16 bgid; >> + >> + /* below is for ring provided buffers */ >> + __u16 buf_nr_pages; >> + __u16 nr_entries; >> + __u32 tail; >> + __u32 mask; >> }; >> struct io_buffer { >> @@ -804,6 +821,7 @@ enum { >> REQ_F_NEED_CLEANUP_BIT, >> REQ_F_POLLED_BIT, >> REQ_F_BUFFER_SELECTED_BIT, >> + REQ_F_BUFFER_RING_BIT, >> REQ_F_COMPLETE_INLINE_BIT, >> REQ_F_REISSUE_BIT, >> REQ_F_CREDS_BIT, >> @@ -855,6 +873,8 @@ enum { >> REQ_F_POLLED = BIT(REQ_F_POLLED_BIT), >> /* buffer already selected */ >> REQ_F_BUFFER_SELECTED = BIT(REQ_F_BUFFER_SELECTED_BIT), >> + /* buffer selected from ring, needs commit */ >> + REQ_F_BUFFER_RING = BIT(REQ_F_BUFFER_RING_BIT), >> /* completion is deferred through io_comp_state */ >> REQ_F_COMPLETE_INLINE = BIT(REQ_F_COMPLETE_INLINE_BIT), >> /* caller should reissue async */ >> @@ -979,6 +999,12 @@ struct io_kiocb { >> /* stores selected buf, valid IFF REQ_F_BUFFER_SELECTED is set */ >> struct io_buffer *kbuf; >> + >> + /* >> + * stores buffer ID for ring provided buffers, valid IFF >> + * REQ_F_BUFFER_RING is set. >> + */ >> + struct io_buffer_list *buf_list; >> }; >> union { >> @@ -1470,8 +1496,14 @@ static inline void io_req_set_rsrc_node(struct io_kiocb *req, >> static unsigned int __io_put_kbuf(struct io_kiocb *req, struct list_head *list) >> { >> - req->flags &= ~REQ_F_BUFFER_SELECTED; >> - list_add(&req->kbuf->list, list); >> + if (req->flags & REQ_F_BUFFER_RING) { >> + if (req->buf_list) >> + req->buf_list->tail++; > > This confused me for some time..seems [tail, head) is the registered > bufs that kernel space can leverage? similar to what pipe logic does. > how about swaping the name of head and tail, this way setting the kernel > as a consumer. But this is just my personal preference.. No agree, I'll make that change. That matches the sq ring as well, which is the same user producer, kernel consumer setup. >> + tail &= bl->mask; >> + if (tail < IO_BUFFER_LIST_BUF_PER_PAGE) { >> + buf = &br->bufs[tail]; >> + } else { >> + int off = tail & (IO_BUFFER_LIST_BUF_PER_PAGE - 1); >> + int index = tail / IO_BUFFER_LIST_BUF_PER_PAGE - 1; > > Could we do some bitwise trick with some compiler check there since for > now IO_BUFFER_LIST_BUF_PER_PAGE is a power of 2. This is known at compile time, so the compiler should already be doing that as it's a constant. >> + buf = page_address(bl->buf_pages[index]); >> + buf += off; >> + } > > I'm not familiar with this part, allow me to ask, is this if else > statement for efficiency? why choose one page as the dividing line We need to index at the right page granularity.
On 5/17/22 23:46, Jens Axboe wrote: > On 5/17/22 8:18 AM, Hao Xu wrote: >> Hi All, >> >> On 5/17/22 00:21, Jens Axboe wrote: >>> Provided buffers allow an application to supply io_uring with buffers >>> that can then be grabbed for a read/receive request, when the data >>> source is ready to deliver data. The existing scheme relies on using >>> IORING_OP_PROVIDE_BUFFERS to do that, but it can be difficult to use >>> in real world applications. It's pretty efficient if the application >>> is able to supply back batches of provided buffers when they have been >>> consumed and the application is ready to recycle them, but if >>> fragmentation occurs in the buffer space, it can become difficult to >>> supply enough buffers at the time. This hurts efficiency. >>> >>> Add a register op, IORING_REGISTER_PBUF_RING, which allows an application >>> to setup a shared queue for each buffer group of provided buffers. The >>> application can then supply buffers simply by adding them to this ring, >>> and the kernel can consume then just as easily. The ring shares the head >>> with the application, the tail remains private in the kernel. >>> >>> Provided buffers setup with IORING_REGISTER_PBUF_RING cannot use >>> IORING_OP_{PROVIDE,REMOVE}_BUFFERS for adding or removing entries to the >>> ring, they must use the mapped ring. Mapped provided buffer rings can >>> co-exist with normal provided buffers, just not within the same group ID. >>> >>> To gauge overhead of the existing scheme and evaluate the mapped ring >>> approach, a simple NOP benchmark was written. It uses a ring of 128 >>> entries, and submits/completes 32 at the time. 'Replenish' is how >>> many buffers are provided back at the time after they have been >>> consumed: >>> >>> Test Replenish NOPs/sec >>> ================================================================ >>> No provided buffers NA ~30M >>> Provided buffers 32 ~16M >>> Provided buffers 1 ~10M >>> Ring buffers 32 ~27M >>> Ring buffers 1 ~27M >>> >>> The ring mapped buffers perform almost as well as not using provided >>> buffers at all, and they don't care if you provided 1 or more back at >>> the same time. This means application can just replenish as they go, >>> rather than need to batch and compact, further reducing overhead in the >>> application. The NOP benchmark above doesn't need to do any compaction, >>> so that overhead isn't even reflected in the above test. >>> >>> Co-developed-by: Dylan Yudaken <dylany@fb.com> >>> Signed-off-by: Jens Axboe <axboe@kernel.dk> >>> --- >>> fs/io_uring.c | 233 ++++++++++++++++++++++++++++++++-- >>> include/uapi/linux/io_uring.h | 36 ++++++ >>> 2 files changed, 257 insertions(+), 12 deletions(-) >>> >>> diff --git a/fs/io_uring.c b/fs/io_uring.c >>> index 5867dcabc73b..776a9f5e5ec7 100644 >>> --- a/fs/io_uring.c >>> +++ b/fs/io_uring.c >>> @@ -285,9 +285,26 @@ struct io_rsrc_data { >>> bool quiesce; >>> }; >>> +#define IO_BUFFER_LIST_BUF_PER_PAGE (PAGE_SIZE / sizeof(struct io_uring_buf)) >>> struct io_buffer_list { >>> - struct list_head buf_list; >>> + /* >>> + * If ->buf_nr_pages is set, then buf_pages/buf_ring are used. If not, >>> + * then these are classic provided buffers and ->buf_list is used. >>> + */ >>> + union { >>> + struct list_head buf_list; >>> + struct { >>> + struct page **buf_pages; >>> + struct io_uring_buf_ring *buf_ring; >>> + }; >>> + }; >>> __u16 bgid; >>> + >>> + /* below is for ring provided buffers */ >>> + __u16 buf_nr_pages; >>> + __u16 nr_entries; >>> + __u32 tail; >>> + __u32 mask; >>> }; >>> struct io_buffer { >>> @@ -804,6 +821,7 @@ enum { >>> REQ_F_NEED_CLEANUP_BIT, >>> REQ_F_POLLED_BIT, >>> REQ_F_BUFFER_SELECTED_BIT, >>> + REQ_F_BUFFER_RING_BIT, >>> REQ_F_COMPLETE_INLINE_BIT, >>> REQ_F_REISSUE_BIT, >>> REQ_F_CREDS_BIT, >>> @@ -855,6 +873,8 @@ enum { >>> REQ_F_POLLED = BIT(REQ_F_POLLED_BIT), >>> /* buffer already selected */ >>> REQ_F_BUFFER_SELECTED = BIT(REQ_F_BUFFER_SELECTED_BIT), >>> + /* buffer selected from ring, needs commit */ >>> + REQ_F_BUFFER_RING = BIT(REQ_F_BUFFER_RING_BIT), >>> /* completion is deferred through io_comp_state */ >>> REQ_F_COMPLETE_INLINE = BIT(REQ_F_COMPLETE_INLINE_BIT), >>> /* caller should reissue async */ >>> @@ -979,6 +999,12 @@ struct io_kiocb { >>> /* stores selected buf, valid IFF REQ_F_BUFFER_SELECTED is set */ >>> struct io_buffer *kbuf; >>> + >>> + /* >>> + * stores buffer ID for ring provided buffers, valid IFF >>> + * REQ_F_BUFFER_RING is set. >>> + */ >>> + struct io_buffer_list *buf_list; >>> }; >>> union { >>> @@ -1470,8 +1496,14 @@ static inline void io_req_set_rsrc_node(struct io_kiocb *req, >>> static unsigned int __io_put_kbuf(struct io_kiocb *req, struct list_head *list) >>> { >>> - req->flags &= ~REQ_F_BUFFER_SELECTED; >>> - list_add(&req->kbuf->list, list); >>> + if (req->flags & REQ_F_BUFFER_RING) { >>> + if (req->buf_list) >>> + req->buf_list->tail++; >> >> This confused me for some time..seems [tail, head) is the registered >> bufs that kernel space can leverage? similar to what pipe logic does. >> how about swaping the name of head and tail, this way setting the kernel >> as a consumer. But this is just my personal preference.. > > No agree, I'll make that change. That matches the sq ring as well, which > is the same user producer, kernel consumer setup. > >>> + tail &= bl->mask; >>> + if (tail < IO_BUFFER_LIST_BUF_PER_PAGE) { >>> + buf = &br->bufs[tail]; >>> + } else { >>> + int off = tail & (IO_BUFFER_LIST_BUF_PER_PAGE - 1); >>> + int index = tail / IO_BUFFER_LIST_BUF_PER_PAGE - 1; >> >> Could we do some bitwise trick with some compiler check there since for >> now IO_BUFFER_LIST_BUF_PER_PAGE is a power of 2. > > This is known at compile time, so the compiler should already be doing > that as it's a constant. > >>> + buf = page_address(bl->buf_pages[index]); >>> + buf += off; >>> + } >> >> I'm not familiar with this part, allow me to ask, is this if else >> statement for efficiency? why choose one page as the dividing line > > We need to index at the right page granularity. Sorry, I didn't get it, why can't we just do buf = &br->bufs[tail]; It seems something is beyond my knowledge.. >
On 5/18/22 4:50 AM, Hao Xu wrote: >> This is known at compile time, so the compiler should already be doing >> that as it's a constant. >> >>>> + buf = page_address(bl->buf_pages[index]); >>>> + buf += off; >>>> + } >>> >>> I'm not familiar with this part, allow me to ask, is this if else >>> statement for efficiency? why choose one page as the dividing line >> >> We need to index at the right page granularity. > > Sorry, I didn't get it, why can't we just do buf = &br->bufs[tail]; > It seems something is beyond my knowledge.. The pages might not be contigious, we have to index from the right page.
diff --git a/fs/io_uring.c b/fs/io_uring.c index 5867dcabc73b..776a9f5e5ec7 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -285,9 +285,26 @@ struct io_rsrc_data { bool quiesce; }; +#define IO_BUFFER_LIST_BUF_PER_PAGE (PAGE_SIZE / sizeof(struct io_uring_buf)) struct io_buffer_list { - struct list_head buf_list; + /* + * If ->buf_nr_pages is set, then buf_pages/buf_ring are used. If not, + * then these are classic provided buffers and ->buf_list is used. + */ + union { + struct list_head buf_list; + struct { + struct page **buf_pages; + struct io_uring_buf_ring *buf_ring; + }; + }; __u16 bgid; + + /* below is for ring provided buffers */ + __u16 buf_nr_pages; + __u16 nr_entries; + __u32 tail; + __u32 mask; }; struct io_buffer { @@ -804,6 +821,7 @@ enum { REQ_F_NEED_CLEANUP_BIT, REQ_F_POLLED_BIT, REQ_F_BUFFER_SELECTED_BIT, + REQ_F_BUFFER_RING_BIT, REQ_F_COMPLETE_INLINE_BIT, REQ_F_REISSUE_BIT, REQ_F_CREDS_BIT, @@ -855,6 +873,8 @@ enum { REQ_F_POLLED = BIT(REQ_F_POLLED_BIT), /* buffer already selected */ REQ_F_BUFFER_SELECTED = BIT(REQ_F_BUFFER_SELECTED_BIT), + /* buffer selected from ring, needs commit */ + REQ_F_BUFFER_RING = BIT(REQ_F_BUFFER_RING_BIT), /* completion is deferred through io_comp_state */ REQ_F_COMPLETE_INLINE = BIT(REQ_F_COMPLETE_INLINE_BIT), /* caller should reissue async */ @@ -979,6 +999,12 @@ struct io_kiocb { /* stores selected buf, valid IFF REQ_F_BUFFER_SELECTED is set */ struct io_buffer *kbuf; + + /* + * stores buffer ID for ring provided buffers, valid IFF + * REQ_F_BUFFER_RING is set. + */ + struct io_buffer_list *buf_list; }; union { @@ -1470,8 +1496,14 @@ static inline void io_req_set_rsrc_node(struct io_kiocb *req, static unsigned int __io_put_kbuf(struct io_kiocb *req, struct list_head *list) { - req->flags &= ~REQ_F_BUFFER_SELECTED; - list_add(&req->kbuf->list, list); + if (req->flags & REQ_F_BUFFER_RING) { + if (req->buf_list) + req->buf_list->tail++; + req->flags &= ~REQ_F_BUFFER_RING; + } else { + list_add(&req->kbuf->list, list); + req->flags &= ~REQ_F_BUFFER_SELECTED; + } return IORING_CQE_F_BUFFER | (req->buf_index << IORING_CQE_BUFFER_SHIFT); } @@ -1480,7 +1512,7 @@ static inline unsigned int io_put_kbuf_comp(struct io_kiocb *req) { lockdep_assert_held(&req->ctx->completion_lock); - if (likely(!(req->flags & REQ_F_BUFFER_SELECTED))) + if (!(req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING))) return 0; return __io_put_kbuf(req, &req->ctx->io_buffers_comp); } @@ -1490,7 +1522,7 @@ static inline unsigned int io_put_kbuf(struct io_kiocb *req, { unsigned int cflags; - if (likely(!(req->flags & REQ_F_BUFFER_SELECTED))) + if (!(req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING))) return 0; /* @@ -1505,7 +1537,10 @@ static inline unsigned int io_put_kbuf(struct io_kiocb *req, * We migrate buffers from the comp_list to the issue cache list * when we need one. */ - if (issue_flags & IO_URING_F_UNLOCKED) { + if (req->flags & REQ_F_BUFFER_RING) { + /* no buffers to recycle for this case */ + cflags = __io_put_kbuf(req, NULL); + } else if (issue_flags & IO_URING_F_UNLOCKED) { struct io_ring_ctx *ctx = req->ctx; spin_lock(&ctx->completion_lock); @@ -1535,11 +1570,23 @@ static void io_kbuf_recycle(struct io_kiocb *req, unsigned issue_flags) struct io_buffer_list *bl; struct io_buffer *buf; - if (likely(!(req->flags & REQ_F_BUFFER_SELECTED))) + if (!(req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING))) return; /* don't recycle if we already did IO to this buffer */ if (req->flags & REQ_F_PARTIAL_IO) return; + /* + * We don't need to recycle for REQ_F_BUFFER_RING, we can just clear + * the flag and hence ensure that bl->tail doesn't get incremented. + * If the tail has already been incremented, hang on to it. + */ + if (req->flags & REQ_F_BUFFER_RING) { + if (req->buf_list) { + req->buf_index = req->buf_list->bgid; + req->flags &= ~REQ_F_BUFFER_RING; + } + return; + } io_ring_submit_lock(ctx, issue_flags); @@ -3487,6 +3534,53 @@ static void __user *io_provided_buffer_select(struct io_kiocb *req, size_t *len, return ret; } +static void __user *io_ring_buffer_select(struct io_kiocb *req, size_t *len, + struct io_buffer_list *bl, + unsigned int issue_flags) +{ + struct io_uring_buf_ring *br = bl->buf_ring; + struct io_uring_buf *buf; + __u32 tail = bl->tail; + + if (unlikely(smp_load_acquire(&br->head) == tail)) { + io_ring_submit_unlock(req->ctx, issue_flags); + return ERR_PTR(-ENOBUFS); + } + + tail &= bl->mask; + if (tail < IO_BUFFER_LIST_BUF_PER_PAGE) { + buf = &br->bufs[tail]; + } else { + int off = tail & (IO_BUFFER_LIST_BUF_PER_PAGE - 1); + int index = tail / IO_BUFFER_LIST_BUF_PER_PAGE - 1; + buf = page_address(bl->buf_pages[index]); + buf += off; + } + if (*len > buf->len) + *len = buf->len; + req->flags |= REQ_F_BUFFER_RING; + req->buf_list = bl; + req->buf_index = buf->bid; + + if (!(issue_flags & IO_URING_F_UNLOCKED)) + return u64_to_user_ptr(buf->addr); + + /* + * If we came in unlocked, we have no choice but to + * consume the buffer here. This does mean it'll be + * pinned until the IO completes. But coming in + * unlocked means we're in io-wq context, hence there + * should be no further retry. For the locked case, the + * caller must ensure to call the commit when the + * transfer completes (or if we get -EAGAIN and must + * poll or retry). + */ + req->buf_list = NULL; + bl->tail++; + io_ring_submit_unlock(req->ctx, issue_flags); + return u64_to_user_ptr(buf->addr); +} + static void __user *io_buffer_select(struct io_kiocb *req, size_t *len, unsigned int issue_flags) { @@ -3502,6 +3596,9 @@ static void __user *io_buffer_select(struct io_kiocb *req, size_t *len, } /* selection helpers drop the submit lock again, if needed */ + if (bl->buf_nr_pages) + return io_ring_buffer_select(req, len, bl, issue_flags); + return io_provided_buffer_select(req, len, bl, issue_flags); } @@ -3558,7 +3655,7 @@ static ssize_t __io_iov_buffer_select(struct io_kiocb *req, struct iovec *iov, static ssize_t io_iov_buffer_select(struct io_kiocb *req, struct iovec *iov, unsigned int issue_flags) { - if (req->flags & REQ_F_BUFFER_SELECTED) { + if (req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING)) { iov[0].iov_base = u64_to_user_ptr(req->rw.addr); iov[0].iov_len = req->rw.len; return 0; @@ -3578,7 +3675,7 @@ static inline bool io_do_buffer_select(struct io_kiocb *req) { if (!(req->flags & REQ_F_BUFFER_SELECT)) return false; - return !(req->flags & REQ_F_BUFFER_SELECTED); + return !(req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING)); } static struct iovec *__io_import_iovec(int rw, struct io_kiocb *req, @@ -4872,6 +4969,17 @@ static int __io_remove_buffers(struct io_ring_ctx *ctx, if (!nbufs) return 0; + if (bl->buf_nr_pages) { + int j; + + for (j = 0; j < bl->buf_nr_pages; j++) + unpin_user_page(bl->buf_pages[j]); + kvfree(bl->buf_pages); + bl->buf_pages = NULL; + bl->buf_nr_pages = 0; + return bl->buf_ring->head - bl->tail; + } + /* the head kbuf is the list itself */ while (!list_empty(&bl->buf_list)) { struct io_buffer *nxt; @@ -4898,8 +5006,12 @@ static int io_remove_buffers(struct io_kiocb *req, unsigned int issue_flags) ret = -ENOENT; bl = io_buffer_get_list(ctx, p->bgid); - if (bl) - ret = __io_remove_buffers(ctx, bl, p->nbufs); + if (bl) { + ret = -EINVAL; + /* can't use provide/remove buffers command on mapped buffers */ + if (!bl->buf_nr_pages) + ret = __io_remove_buffers(ctx, bl, p->nbufs); + } if (ret < 0) req_set_fail(req); @@ -5047,7 +5159,7 @@ static int io_provide_buffers(struct io_kiocb *req, unsigned int issue_flags) bl = io_buffer_get_list(ctx, p->bgid); if (unlikely(!bl)) { - bl = kmalloc(sizeof(*bl), GFP_KERNEL); + bl = kzalloc(sizeof(*bl), GFP_KERNEL); if (!bl) { ret = -ENOMEM; goto err; @@ -5058,6 +5170,11 @@ static int io_provide_buffers(struct io_kiocb *req, unsigned int issue_flags) goto err; } } + /* can't add buffers via this command for a mapped buffer ring */ + if (bl->buf_nr_pages) { + ret = -EINVAL; + goto err; + } ret = io_add_buffers(ctx, p, bl); err: @@ -12011,6 +12128,83 @@ static __cold int io_register_iowq_max_workers(struct io_ring_ctx *ctx, return ret; } +static int io_register_pbuf_ring(struct io_ring_ctx *ctx, void __user *arg) +{ + struct io_uring_buf_ring *br; + struct io_uring_buf_reg reg; + struct io_buffer_list *bl; + struct page **pages; + int nr_pages; + + if (copy_from_user(®, arg, sizeof(reg))) + return -EFAULT; + + if (reg.pad || reg.resv[0] || reg.resv[1] || reg.resv[2]) + return -EINVAL; + if (!reg.ring_addr) + return -EFAULT; + if (reg.ring_addr & ~PAGE_MASK) + return -EINVAL; + if (!is_power_of_2(reg.ring_entries)) + return -EINVAL; + + if (unlikely(reg.bgid < BGID_ARRAY && !ctx->io_bl)) { + int ret = io_init_bl_list(ctx); + if (ret) + return ret; + } + + bl = io_buffer_get_list(ctx, reg.bgid); + if (bl && bl->buf_nr_pages) + return -EEXIST; + if (!bl) { + bl = kzalloc(sizeof(*bl), GFP_KERNEL); + if (!bl) + return -ENOMEM; + } + + pages = io_pin_pages(reg.ring_addr, + struct_size(br, bufs, reg.ring_entries), + &nr_pages); + if (IS_ERR(pages)) { + kfree(bl); + return PTR_ERR(pages); + } + + br = page_address(pages[0]); + bl->buf_pages = pages; + bl->buf_nr_pages = nr_pages; + bl->nr_entries = reg.ring_entries; + bl->buf_ring = br; + bl->mask = reg.ring_entries - 1; + io_buffer_add_list(ctx, bl, reg.bgid); + return 0; +} + +static int io_unregister_pbuf_ring(struct io_ring_ctx *ctx, void __user *arg) +{ + struct io_uring_buf_reg reg; + struct io_buffer_list *bl; + + if (copy_from_user(®, arg, sizeof(reg))) + return -EFAULT; + if (reg.pad || reg.resv[0] || reg.resv[1] || reg.resv[2]) + return -EINVAL; + + bl = io_buffer_get_list(ctx, reg.bgid); + if (!bl) + return -ENOENT; + if (!bl->buf_nr_pages) + return -EINVAL; + + __io_remove_buffers(ctx, bl, -1U); + if (bl->bgid >= BGID_ARRAY) { + xa_erase(&ctx->io_bl_xa, bl->bgid); + kfree(bl); + } + return 0; +} + static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode, void __user *arg, unsigned nr_args) __releases(ctx->uring_lock) @@ -12142,6 +12336,18 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode, case IORING_UNREGISTER_RING_FDS: ret = io_ringfd_unregister(ctx, arg, nr_args); break; + case IORING_REGISTER_PBUF_RING: + ret = -EINVAL; + if (!arg || nr_args != 1) + break; + ret = io_register_pbuf_ring(ctx, arg); + break; + case IORING_UNREGISTER_PBUF_RING: + ret = -EINVAL; + if (!arg || nr_args != 1) + break; + ret = io_unregister_pbuf_ring(ctx, arg); + break; default: ret = -EINVAL; break; @@ -12227,6 +12433,9 @@ static int __init io_uring_init(void) /* ->buf_index is u16 */ BUILD_BUG_ON(IORING_MAX_REG_BUFFERS >= (1u << 16)); BUILD_BUG_ON(BGID_ARRAY * sizeof(struct io_buffer_list) > PAGE_SIZE); + BUILD_BUG_ON(offsetof(struct io_uring_buf_ring, bufs) != 0); + BUILD_BUG_ON(offsetof(struct io_uring_buf, resv) != + offsetof(struct io_uring_buf_ring, head)); /* should fit into one byte */ BUILD_BUG_ON(SQE_VALID_FLAGS >= (1 << 8)); diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h index 15f821af9242..90d78428317a 100644 --- a/include/uapi/linux/io_uring.h +++ b/include/uapi/linux/io_uring.h @@ -384,6 +384,10 @@ enum { IORING_REGISTER_RING_FDS = 20, IORING_UNREGISTER_RING_FDS = 21, + /* register ring based provide buffer group */ + IORING_REGISTER_PBUF_RING = 22, + IORING_UNREGISTER_PBUF_RING = 23, + /* this goes last */ IORING_REGISTER_LAST }; @@ -461,6 +465,38 @@ struct io_uring_restriction { __u32 resv2[3]; }; +struct io_uring_buf { + __u64 addr; + __u32 len; + __u16 bid; + __u16 resv; +}; + +struct io_uring_buf_ring { + union { + /* + * To avoid spilling into more pages than we need to, the + * ring head is overlaid with the io_uring_buf->resv field. + */ + struct { + __u64 resv1; + __u32 resv2; + __u16 resv3; + __u16 head; + }; + struct io_uring_buf bufs[0]; + }; +}; + +/* argument for IORING_(UN)REGISTER_PBUF_RING */ +struct io_uring_buf_reg { + __u64 ring_addr; + __u32 ring_entries; + __u16 bgid; + __u16 pad; + __u64 resv[3]; +}; + /* * io_uring_restriction->opcode values */
Provided buffers allow an application to supply io_uring with buffers that can then be grabbed for a read/receive request, when the data source is ready to deliver data. The existing scheme relies on using IORING_OP_PROVIDE_BUFFERS to do that, but it can be difficult to use in real world applications. It's pretty efficient if the application is able to supply back batches of provided buffers when they have been consumed and the application is ready to recycle them, but if fragmentation occurs in the buffer space, it can become difficult to supply enough buffers at the time. This hurts efficiency. Add a register op, IORING_REGISTER_PBUF_RING, which allows an application to setup a shared queue for each buffer group of provided buffers. The application can then supply buffers simply by adding them to this ring, and the kernel can consume then just as easily. The ring shares the head with the application, the tail remains private in the kernel. Provided buffers setup with IORING_REGISTER_PBUF_RING cannot use IORING_OP_{PROVIDE,REMOVE}_BUFFERS for adding or removing entries to the ring, they must use the mapped ring. Mapped provided buffer rings can co-exist with normal provided buffers, just not within the same group ID. To gauge overhead of the existing scheme and evaluate the mapped ring approach, a simple NOP benchmark was written. It uses a ring of 128 entries, and submits/completes 32 at the time. 'Replenish' is how many buffers are provided back at the time after they have been consumed: Test Replenish NOPs/sec ================================================================ No provided buffers NA ~30M Provided buffers 32 ~16M Provided buffers 1 ~10M Ring buffers 32 ~27M Ring buffers 1 ~27M The ring mapped buffers perform almost as well as not using provided buffers at all, and they don't care if you provided 1 or more back at the same time. This means application can just replenish as they go, rather than need to batch and compact, further reducing overhead in the application. The NOP benchmark above doesn't need to do any compaction, so that overhead isn't even reflected in the above test. Co-developed-by: Dylan Yudaken <dylany@fb.com> Signed-off-by: Jens Axboe <axboe@kernel.dk> --- fs/io_uring.c | 233 ++++++++++++++++++++++++++++++++-- include/uapi/linux/io_uring.h | 36 ++++++ 2 files changed, 257 insertions(+), 12 deletions(-)