[dpdk-dev] [PATCH v4 0/2] Add VHOST PMD
On 2015/11/17 22:26, Yuanhan Liu wrote: > On Fri, Nov 13, 2015 at 03:50:16PM +0900, Tetsuya Mukawa wrote: >> On 2015/11/13 14:32, Yuanhan Liu wrote: >>> On Fri, Nov 13, 2015 at 02:20:29PM +0900, Tetsuya Mukawa wrote: The patch introduces a new PMD. This PMD is implemented as thin wrapper of librte_vhost. * Known issue. We may see issues while handling RESET_OWNER message. These handlings are done in vhost library, so not a part of vhost PMD. So far, we are waiting for QEMU fixing. >>> Fix patches have already been applied. Please help test :) >>> >>> --yliu >> Hi Yuanhan, >> >> It seems there might be an another issue related with "vq->callfd" in >> vhost library. >> We may miss something to handle the value correctly. >> >> Anyway, here are steps. >> 1. Apply vhost PMD patch. >> (I guess you don't need it to reproduce the issue, but to reproduce it, >> using the PMD may be easy) >> 2. Start testpmd on host with vhost-user PMD. >> 3. Start QEMU with virtio-net device. >> 4. Login QEMU. >> 5. Bind the virtio-net device to igb_uio. >> 6. Start testpmd in QEMU. >> 7. Quit testmd in QEMU. >> 8. Start testpmd again in QEMU. >> >> It seems when last command is executed, testpmd on host doesn't receive >> SET_VRING_CALL message from QEMU. >> Because of this, testpmd on host assumes virtio-net device is not ready. >> (I made sure virtio_is_ready() was failed on host). >> >> According to QEMU source code, SET_VRING_KICK will be called when >> virtqueue starts, but SET_VRING_CALL will be called when virtqueue is >> initialized. >> Not sure exactly, might be "vq->call" will be valid while connection is >> established? > Yes, it would be valid as far as we don't reset it from another > set_vring_call. So, we should not reset it on reset_device(). > > --yliu Hi Yuanhan, Thanks for checking. I will submit the patch for this today. Tetsuya >> Also I've found a workaround. >> Please execute after step7. >> >> 8. Bind the virtio-net device to virtio-pci kernel driver. >> 9. Bind the virtio-net device to igb_uio. >> 10. Start testpmd in QEMU. >> >> When step8 is executed, connection will be re-established, and testpmd >> on host will be able to receive SET_VRING_CALL. >> Then testpmd on host can start. >> >> Thanks, >> Tetsuya
[dpdk-dev] [PATCH v4 0/2] Add VHOST PMD
On Fri, Nov 13, 2015 at 03:50:16PM +0900, Tetsuya Mukawa wrote: > On 2015/11/13 14:32, Yuanhan Liu wrote: > > On Fri, Nov 13, 2015 at 02:20:29PM +0900, Tetsuya Mukawa wrote: > >> The patch introduces a new PMD. This PMD is implemented as thin wrapper > >> of librte_vhost. > >> > >> * Known issue. > >> We may see issues while handling RESET_OWNER message. > >> These handlings are done in vhost library, so not a part of vhost PMD. > >> So far, we are waiting for QEMU fixing. > > Fix patches have already been applied. Please help test :) > > > > --yliu > > Hi Yuanhan, > > It seems there might be an another issue related with "vq->callfd" in > vhost library. > We may miss something to handle the value correctly. > > Anyway, here are steps. > 1. Apply vhost PMD patch. > (I guess you don't need it to reproduce the issue, but to reproduce it, > using the PMD may be easy) > 2. Start testpmd on host with vhost-user PMD. > 3. Start QEMU with virtio-net device. > 4. Login QEMU. > 5. Bind the virtio-net device to igb_uio. > 6. Start testpmd in QEMU. > 7. Quit testmd in QEMU. > 8. Start testpmd again in QEMU. > > It seems when last command is executed, testpmd on host doesn't receive > SET_VRING_CALL message from QEMU. > Because of this, testpmd on host assumes virtio-net device is not ready. > (I made sure virtio_is_ready() was failed on host). > > According to QEMU source code, SET_VRING_KICK will be called when > virtqueue starts, but SET_VRING_CALL will be called when virtqueue is > initialized. > Not sure exactly, might be "vq->call" will be valid while connection is > established? Yes, it would be valid as far as we don't reset it from another set_vring_call. So, we should not reset it on reset_device(). --yliu > > Also I've found a workaround. > Please execute after step7. > > 8. Bind the virtio-net device to virtio-pci kernel driver. > 9. Bind the virtio-net device to igb_uio. > 10. Start testpmd in QEMU. > > When step8 is executed, connection will be re-established, and testpmd > on host will be able to receive SET_VRING_CALL. > Then testpmd on host can start. > > Thanks, > Tetsuya
[dpdk-dev] [PATCH v4 0/2] Add VHOST PMD
On 2015/11/13 14:32, Yuanhan Liu wrote: > On Fri, Nov 13, 2015 at 02:20:29PM +0900, Tetsuya Mukawa wrote: >> The patch introduces a new PMD. This PMD is implemented as thin wrapper >> of librte_vhost. >> >> * Known issue. >> We may see issues while handling RESET_OWNER message. >> These handlings are done in vhost library, so not a part of vhost PMD. >> So far, we are waiting for QEMU fixing. > Fix patches have already been applied. Please help test :) > > --yliu Hi Yuanhan, It seems there might be an another issue related with "vq->callfd" in vhost library. We may miss something to handle the value correctly. Anyway, here are steps. 1. Apply vhost PMD patch. (I guess you don't need it to reproduce the issue, but to reproduce it, using the PMD may be easy) 2. Start testpmd on host with vhost-user PMD. 3. Start QEMU with virtio-net device. 4. Login QEMU. 5. Bind the virtio-net device to igb_uio. 6. Start testpmd in QEMU. 7. Quit testmd in QEMU. 8. Start testpmd again in QEMU. It seems when last command is executed, testpmd on host doesn't receive SET_VRING_CALL message from QEMU. Because of this, testpmd on host assumes virtio-net device is not ready. (I made sure virtio_is_ready() was failed on host). According to QEMU source code, SET_VRING_KICK will be called when virtqueue starts, but SET_VRING_CALL will be called when virtqueue is initialized. Not sure exactly, might be "vq->call" will be valid while connection is established? Also I've found a workaround. Please execute after step7. 8. Bind the virtio-net device to virtio-pci kernel driver. 9. Bind the virtio-net device to igb_uio. 10. Start testpmd in QEMU. When step8 is executed, connection will be re-established, and testpmd on host will be able to receive SET_VRING_CALL. Then testpmd on host can start. Thanks, Tetsuya
[dpdk-dev] [PATCH v4 0/2] Add VHOST PMD
On 2015/11/13 14:32, Yuanhan Liu wrote: > On Fri, Nov 13, 2015 at 02:20:29PM +0900, Tetsuya Mukawa wrote: >> The patch introduces a new PMD. This PMD is implemented as thin wrapper >> of librte_vhost. >> >> * Known issue. >> We may see issues while handling RESET_OWNER message. >> These handlings are done in vhost library, so not a part of vhost PMD. >> So far, we are waiting for QEMU fixing. > Fix patches have already been applied. Please help test :) > > --yliu Thanks! I have checked it, and it worked! Tetsuya
[dpdk-dev] [PATCH v4 0/2] Add VHOST PMD
The patch introduces a new PMD. This PMD is implemented as thin wrapper of librte_vhost. * Known issue. We may see issues while handling RESET_OWNER message. These handlings are done in vhost library, so not a part of vhost PMD. So far, we are waiting for QEMU fixing. PATCH v4 changes: - Rebase on latest DPDK tree. - Fix cording style. - Fix code not to invoke multiple messaging handling threads. - Fix code to handle vdev parameters correctly. - Remove needless cast. - Remove needless if-condition before rt_free(). PATCH v3 changes: - Rebase on latest matser - Specify correct queue_id in RX/TX function. PATCH v2 changes: - Remove a below patch that fixes vhost library. The patch was applied as a separate patch. - vhost: fix crash with multiqueue enabled - Fix typos. (Thanks to Thomas, Monjalon) - Rebase on latest tree with above bernard's patches. PATCH v1 changes: - Support vhost multiple queues. - Rebase on "remove pci driver from vdevs". - Optimize RX/TX functions. - Fix resource leaks. - Fix compile issue. - Add patch to fix vhost library. RFC PATCH v3 changes: - Optimize performance. In RX/TX functions, change code to access only per core data. - Add below API to allow user to use vhost library APIs for a port managed by vhost PMD. There are a few limitations. See "rte_eth_vhost.h". - rte_eth_vhost_portid2vdev() To support this functionality, vhost library is also changed. Anyway, if users doesn't use vhost PMD, can fully use vhost library APIs. - Add code to support vhost multiple queues. Actually, multiple queues functionality is not enabled so far. RFC PATCH v2 changes: - Fix issues reported by checkpatch.pl (Thanks to Stephen Hemminger) Tetsuya Mukawa (2): vhost: Add callback and private data for vhost PMD vhost: Add VHOST PMD config/common_linuxapp| 6 + doc/guides/nics/index.rst | 1 + doc/guides/rel_notes/release_2_2.rst | 2 + drivers/net/Makefile | 4 + drivers/net/vhost/Makefile| 62 ++ drivers/net/vhost/rte_eth_vhost.c | 783 ++ drivers/net/vhost/rte_eth_vhost.h | 65 +++ drivers/net/vhost/rte_pmd_vhost_version.map | 8 + lib/librte_vhost/rte_vhost_version.map| 6 + lib/librte_vhost/rte_virtio_net.h | 3 + lib/librte_vhost/vhost_user/virtio-net-user.c | 13 +- lib/librte_vhost/virtio-net.c | 60 +- lib/librte_vhost/virtio-net.h | 4 +- mk/rte.app.mk | 8 +- 14 files changed, 1011 insertions(+), 14 deletions(-) create mode 100644 drivers/net/vhost/Makefile create mode 100644 drivers/net/vhost/rte_eth_vhost.c create mode 100644 drivers/net/vhost/rte_eth_vhost.h create mode 100644 drivers/net/vhost/rte_pmd_vhost_version.map -- 2.1.4
[dpdk-dev] [PATCH v4 0/2] Add VHOST PMD
On Fri, Nov 13, 2015 at 02:20:29PM +0900, Tetsuya Mukawa wrote: > The patch introduces a new PMD. This PMD is implemented as thin wrapper > of librte_vhost. > > * Known issue. > We may see issues while handling RESET_OWNER message. > These handlings are done in vhost library, so not a part of vhost PMD. > So far, we are waiting for QEMU fixing. Fix patches have already been applied. Please help test :) --yliu
