All of lore.kernel.org
 help / color / mirror / Atom feed
From: marcandre.lureau@redhat.com
To: qemu-devel@nongnu.org
Cc: mukawa@igel.co.jp, yuanhan.liu@linux.intel.com,
	victork@redhat.com, jonshin@cisco.com, mst@redhat.com,
	"Marc-André Lureau" <marcandre.lureau@redhat.com>
Subject: [Qemu-devel] [PATCH v6 00/33] vhost-user reconnect fixes
Date: Wed, 27 Jul 2016 01:14:54 +0400	[thread overview]
Message-ID: <20160726211527.28510-1-marcandre.lureau@redhat.com> (raw)

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Hi,

Since 'vhost-user: simple reconnection support' has been merged, it is
possible to disconnect and reconnect a vhost-user backend. However,
many code paths in qemu may trigger assert() when the backend is
disconnected.

There are also code paths that are wrong, see "don't assume opaque is
a fd" patch for an example. Once those patches are reviewed & merged,
they are good candidates for stable too.

Some assert() could simply be replaced by error_report() or silently
fail since they are recoverable cases. Some missing error checks can
also help prevent later issues. The errors are reported up to vhost.c,
as the vhost-user backend alone doesn't handle disconnected state
transparently so far. There are still problematic code paths left
after this series, for example, starting a migration with a
disconnected backend will abort(). It is likely that other problematic
code path exists (vhost_scsi_start failure is fatal, but there are no
vhost-user backend that I know yet).

In many cases, the code assumes get_vhost_net() will be non-NULL after
a succesful connection, so I changed it to stay after a disconnect
until the new connection comes (as suggested by Michael).

Since there is feature checks on reconnection, qemu should wait for
the initial connection feature negotiation to complete. The test added
demonstrates this. Additionally, a regression was found during v2,
which could have been spotted with a multiqueue test, so add a basic
one that would have exhibited the issue.

For convenience, the series is also available on:
https://github.com/elmarco/qemu, branch vhost-user-reconnect

v6:
- fixes after Ilya Maximets review
- add "disconnect on HUP" for cases where peer disconnect doesn't
  trigger qemu disconnect event (reconnect won't work)
- add a flags mismatch on reconnect test

v5:
- rebased
- use a VHOST_OPS_DEBUG macro to print vhost_ops errors
- replace assert(foo != NULL) with assert(foo)
- add "RFC: vhost: do not update last avail idx"

v4:
- change notify_migration_done() patch to be VHOST_BACKEND_TYPE_USER
  specific, to avoid having to handle the case where the backend
  doesn't implement the callback
- change vhost_dev_cleanup() to assert on empty log, instead of
  adding a call to vhost_log_put()
- made "keep vhost_net after a disconnection" more robust, got rid of
  the acked_features field
- improve commit log, and some patch reorganization for clarity

v3:
- add vhost-user multiqueue test, which would have helped to find the
  following fix
- fix waiting on vhost-user connection with multiqueue (found by
  Yuanhan Liu)
- merge vhost_user_{read,write}() error checking patches
- add error reporting to vhost_user_read() (similar to
  vhost_user_write())
- add a vhost_net_set_backend() wrapper to help with potential crash
- some leak fixes

v2:
- some patch ordering: minor fix, close(fd) fix,
  assert/fprintf->error_report, check and return error,
  vhost_dev_cleanup() fixes, keep vhost_net after a disconnect, wait
  until connection is ready
- merge read/write error checks
- do not rely on link state to check vhost-user init completed

Marc-André Lureau (33):
  misc: indentation
  vhost-user: minor simplification
  vhost-user: disconnect on HUP
  vhost: don't assume opaque is a fd, use backend cleanup
  vhost: make vhost_log_put() idempotent
  vhost: assert the log was cleaned up
  vhost: fix cleanup on not fully initialized device
  vhost: make vhost_dev_cleanup() idempotent
  vhost-net: always call vhost_dev_cleanup() on failure
  vhost: fix calling vhost_dev_cleanup() after vhost_dev_init()
  vhost: do not assert() on vhost_ops failure
  vhost: add missing VHOST_OPS_DEBUG
  vhost: use error_report() instead of fprintf(stderr,...)
  qemu-char: fix qemu_chr_fe_set_msgfds() crash when disconnected
  vhost-user: call set_msgfds unconditionally
  vhost-user: check qemu_chr_fe_set_msgfds() return value
  vhost-user: check vhost_user_{read,write}() return value
  vhost-user: keep vhost_net after a disconnection
  vhost-user: add get_vhost_net() assertions
  Revert "vhost-net: do not crash if backend is not present"
  vhost-net: vhost_migration_done is vhost-user specific
  vhost: add assert() to check runtime behaviour
  char: add chr_wait_connected callback
  char: add and use tcp_chr_wait_connected
  vhost-user: wait until backend init is completed
  tests: plug some leaks in virtio-net-test
  tests: fix vhost-user-test leak
  tests: add /vhost-user/connect-fail test
  tests: add a simple /vhost-user/multiqueue test
  vhost-user: add error report in vhost_user_write()
  vhost: add vhost_net_set_backend()
  vhost-user-test: add flags mismatch test
  RFC: vhost: do not update last avail idx on get_vring_base() failure

 hw/net/vhost_net.c        |  34 +++-----
 hw/virtio/vhost-user.c    |  67 ++++++++++-----
 hw/virtio/vhost.c         | 162 +++++++++++++++++++++++-------------
 include/hw/virtio/vhost.h |   4 +
 include/sysemu/char.h     |   8 ++
 net/tap.c                 |   1 +
 net/vhost-user.c          |  65 +++++++++------
 qemu-char.c               |  82 ++++++++++++------
 tests/Makefile.include    |   2 +-
 tests/vhost-user-test.c   | 206 +++++++++++++++++++++++++++++++++++++++++++++-
 tests/virtio-net-test.c   |  12 ++-
 11 files changed, 486 insertions(+), 157 deletions(-)

-- 
2.9.0

             reply	other threads:[~2016-07-26 21:15 UTC|newest]

Thread overview: 36+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-07-26 21:14 marcandre.lureau [this message]
2016-07-26 21:14 ` [Qemu-devel] [PATCH v6 01/33] misc: indentation marcandre.lureau
2016-07-26 21:14 ` [Qemu-devel] [PATCH v6 02/33] vhost-user: minor simplification marcandre.lureau
2016-07-26 21:14 ` [Qemu-devel] [PATCH v6 03/33] vhost-user: disconnect on HUP marcandre.lureau
2016-07-26 21:14 ` [Qemu-devel] [PATCH v6 04/33] vhost: don't assume opaque is a fd, use backend cleanup marcandre.lureau
2016-07-26 21:14 ` [Qemu-devel] [PATCH v6 05/33] vhost: make vhost_log_put() idempotent marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 06/33] vhost: assert the log was cleaned up marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 07/33] vhost: fix cleanup on not fully initialized device marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 08/33] vhost: make vhost_dev_cleanup() idempotent marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 09/33] vhost-net: always call vhost_dev_cleanup() on failure marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 10/33] vhost: fix calling vhost_dev_cleanup() after vhost_dev_init() marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 11/33] vhost: do not assert() on vhost_ops failure marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 12/33] vhost: add missing VHOST_OPS_DEBUG marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 13/33] vhost: use error_report() instead of fprintf(stderr, ...) marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 14/33] qemu-char: fix qemu_chr_fe_set_msgfds() crash when disconnected marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 15/33] vhost-user: call set_msgfds unconditionally marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 16/33] vhost-user: check qemu_chr_fe_set_msgfds() return value marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 17/33] vhost-user: check vhost_user_{read, write}() " marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 18/33] vhost-user: keep vhost_net after a disconnection marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 19/33] vhost-user: add get_vhost_net() assertions marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 20/33] Revert "vhost-net: do not crash if backend is not present" marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 21/33] vhost-net: vhost_migration_done is vhost-user specific marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 22/33] vhost: add assert() to check runtime behaviour marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 23/33] char: add chr_wait_connected callback marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 24/33] char: add and use tcp_chr_wait_connected marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 25/33] vhost-user: wait until backend init is completed marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 26/33] tests: plug some leaks in virtio-net-test marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 27/33] tests: fix vhost-user-test leak marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 28/33] tests: add /vhost-user/connect-fail test marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 29/33] tests: add a simple /vhost-user/multiqueue test marcandre.lureau
2016-07-27  8:26   ` Marc-André Lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 30/33] vhost-user: add error report in vhost_user_write() marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 31/33] vhost: add vhost_net_set_backend() marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 32/33] vhost-user-test: add flags mismatch test marcandre.lureau
2016-07-26 21:15 ` [Qemu-devel] [PATCH v6 33/33] RFC: vhost: do not update last avail idx on get_vring_base() failure marcandre.lureau
2016-07-29  3:17 ` [Qemu-devel] [PATCH v6 00/33] vhost-user reconnect fixes Michael S. Tsirkin

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20160726211527.28510-1-marcandre.lureau@redhat.com \
    --to=marcandre.lureau@redhat.com \
    --cc=jonshin@cisco.com \
    --cc=mst@redhat.com \
    --cc=mukawa@igel.co.jp \
    --cc=qemu-devel@nongnu.org \
    --cc=victork@redhat.com \
    --cc=yuanhan.liu@linux.intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.