Date: Wed, 27 Jul 2016 06:36:58 +0000 (UTC) From: Sepherosa Ziehau <sephe@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r303366 - head/sys/dev/hyperv/vmbus Message-ID: <201607270636.u6R6awvA073664@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: sephe Date: Wed Jul 27 06:36:57 2016 New Revision: 303366 URL: https://svnweb.freebsd.org/changeset/base/303366 Log: hyperv/vmbus: Update comment for bufring MFC after: 1 week Sponsored by: Microsoft Differential Revision: https://reviews.freebsd.org/D7314 Modified: head/sys/dev/hyperv/vmbus/hv_ring_buffer.c Modified: head/sys/dev/hyperv/vmbus/hv_ring_buffer.c ============================================================================== --- head/sys/dev/hyperv/vmbus/hv_ring_buffer.c Wed Jul 27 06:31:40 2016 (r303365) +++ head/sys/dev/hyperv/vmbus/hv_ring_buffer.c Wed Jul 27 06:36:57 2016 (r303366) @@ -144,25 +144,23 @@ vmbus_rxbr_intr_unmask(struct vmbus_rxbr /* * Now check to see if the ring buffer is still empty. * If it is not, we raced and we need to process new - * incoming messages. + * incoming channel packets. */ return vmbus_rxbr_avail(rbr); } /* - * When we write to the ring buffer, check if the host needs to - * be signaled. Here is the details of this protocol: + * When we write to the ring buffer, check if the host needs to be + * signaled. * - * 1. The host guarantees that while it is draining the - * ring buffer, it will set the interrupt_mask to - * indicate it does not need to be interrupted when - * new data is placed. - * - * 2. The host guarantees that it will completely drain - * the ring buffer before exiting the read loop. Further, - * once the ring buffer is empty, it will clear the - * interrupt_mask and re-check to see if new data has - * arrived. + * The contract: + * - The host guarantees that while it is draining the TX bufring, + * it will set the br_imask to indicate it does not need to be + * interrupted when new data are added. + * - The host guarantees that it will completely drain the TX bufring + * before exiting the read loop. Further, once the TX bufring is + * empty, it will clear the br_imask and re-check to see if new + * data have arrived. */ static boolean_t hv_ring_buffer_needsig_on_write(uint32_t old_write_location, @@ -172,8 +170,10 @@ hv_ring_buffer_needsig_on_write(uint32_t if (tbr->txbr_imask) return (FALSE); + /* XXX only compiler fence is needed */ /* Read memory barrier */ rmb(); + /* * This is the only case we need to signal when the * ring transitions from being empty to non-empty. @@ -252,7 +252,6 @@ vmbus_txbr_write(struct vmbus_txbr *tbr, for (i = 0; i < iovlen; i++) total_bytes_to_write += iov[i].iov_len; - total_bytes_to_write += sizeof(uint64_t); mtx_lock_spin(&tbr->txbr_lock); @@ -260,9 +259,11 @@ vmbus_txbr_write(struct vmbus_txbr *tbr, byte_avail_to_write = vmbus_txbr_avail(tbr); /* - * If there is only room for the packet, assume it is full. - * Otherwise, the next time around, we think the ring buffer - * is empty since the read index == write index + * NOTE: + * If this write is going to make br_windex same as br_rindex, + * i.e. the available space for write is same as the write size, + * we can't do it then, since br_windex == br_rindex means that + * the bufring is empty. */ if (byte_avail_to_write <= total_bytes_to_write) { mtx_unlock_spin(&tbr->txbr_lock); @@ -270,7 +271,7 @@ vmbus_txbr_write(struct vmbus_txbr *tbr, } /* - * Write to the ring buffer + * Copy the scattered channel packet to the TX bufring. */ next_write_location = tbr->txbr_windex; @@ -282,20 +283,20 @@ vmbus_txbr_write(struct vmbus_txbr *tbr, } /* - * Set previous packet start + * Set the offset of the current channel packet. */ prev_indices = ((uint64_t)tbr->txbr_windex) << 32; - next_write_location = copy_to_ring_buffer(tbr, next_write_location, (char *)&prev_indices, sizeof(uint64_t)); /* + * XXX only compiler fence is needed. * Full memory barrier before upding the write index. */ mb(); /* - * Now, update the write location + * Now, update the write index. */ tbr->txbr_windex = next_write_location; @@ -315,20 +316,12 @@ vmbus_rxbr_peek(struct vmbus_rxbr *rbr, mtx_lock_spin(&rbr->rxbr_lock); bytesAvailToRead = vmbus_rxbr_avail(rbr); - - /* - * Make sure there is something to read - */ if (bytesAvailToRead < dlen) { mtx_unlock_spin(&rbr->rxbr_lock); return (EAGAIN); } - /* - * Convert to byte offset - */ nextReadLocation = rbr->rxbr_rindex; - nextReadLocation = copy_from_ring_buffer(rbr, data, dlen, nextReadLocation); @@ -349,24 +342,27 @@ vmbus_rxbr_read(struct vmbus_rxbr *rbr, mtx_lock_spin(&rbr->rxbr_lock); bytes_avail_to_read = vmbus_rxbr_avail(rbr); - - /* - * Make sure there is something to read - */ if (bytes_avail_to_read < dlen) { mtx_unlock_spin(&rbr->rxbr_lock); return (EAGAIN); } + /* + * Copy channel packet from RX bufring. + */ next_read_location = (rbr->rxbr_rindex + offset) % rbr->rxbr_dsize; - next_read_location = copy_from_ring_buffer(rbr, data, dlen, next_read_location); + /* + * Discard this channel packet's start offset, which is useless + * for us. + */ next_read_location = copy_from_ring_buffer(rbr, (char *)&prev_indices, sizeof(uint64_t), next_read_location); /* + * XXX only compiler fence is needed. * Make sure all reads are done before we update the read index since * the writer may start writing to the read area once the read index * is updated. @@ -383,11 +379,6 @@ vmbus_rxbr_read(struct vmbus_rxbr *rbr, return (0); } -/** - * @brief Helper routine to copy from source to ring buffer. - * - * Assume there is enough room. Handles wrap-around in dest case only! - */ static uint32_t copy_to_ring_buffer(const struct vmbus_txbr *tbr, uint32_t start_write_offset, const uint8_t *src, uint32_t src_len) @@ -397,7 +388,7 @@ copy_to_ring_buffer(const struct vmbus_t uint32_t fragLen; if (src_len > ring_buffer_size - start_write_offset) { - /* wrap-around detected! */ + /* Wrap-around detected! */ fragLen = ring_buffer_size - start_write_offset; memcpy(ring_buffer + start_write_offset, src, fragLen); memcpy(ring_buffer, src + fragLen, src_len - fragLen); @@ -411,11 +402,6 @@ copy_to_ring_buffer(const struct vmbus_t return (start_write_offset); } -/** - * @brief Helper routine to copy to source from ring buffer. - * - * Assume there is enough room. Handles wrap-around in src case only! - */ static uint32_t copy_from_ring_buffer(const struct vmbus_rxbr *rbr, char *dest, uint32_t dest_len, uint32_t start_read_offset) @@ -425,7 +411,7 @@ copy_from_ring_buffer(const struct vmbus uint32_t ring_buffer_size = rbr->rxbr_dsize; if (dest_len > ring_buffer_size - start_read_offset) { - /* wrap-around detected at the src */ + /* Wrap-around detected at the src */ fragLen = ring_buffer_size - start_read_offset; memcpy(dest, ring_buffer + start_read_offset, fragLen); memcpy(dest + fragLen, ring_buffer, dest_len - fragLen);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201607270636.u6R6awvA073664>