From: Konstantin Osipov <kostja.osipov@gmail.com> To: Georgy Kirichenko <georgy@tarantool.org> Cc: tarantool-patches@dev.tarantool.org Subject: Re: [Tarantool-patches] [PATCH v4 03/11] coio: do not allow parallel usage of coio Date: Thu, 19 Mar 2020 21:09:12 +0300 [thread overview] Message-ID: <20200319180912.GA3662@atlas> (raw) In-Reply-To: <17a0bca9bc72901c99f75bd14913d54de10781bd.1581500169.git.georgy@tarantool.org> * Georgy Kirichenko <georgy@tarantool.org> [20/02/12 13:09]: > Simultaneous usage of one coio from two or more fiber could lead > to undefined behavior as coio routines are replacing awaiting fiber > (a data member) and stopping watcher without any relevance if there > any other users of the coio object. Such behavior could lead to > an applier invalid stream issue #4040. > The proposal is to disable an active coio reuse by returning a fake > EINPROGRESS error. I am not aware of any cases when coio is used by multiple fibers: it's a violation of the coio contract. If you suspect it is violated in some cases, please add the scenario description to the commit comment. Then, to better identify the case, your patch should add an assert in debug mode. For release mode, I suggest we set a more clear error so that it's easy then to spot in the log file and act upon, not EINPROGRESS. Something like ERR_OPEN_A_BUG. > > Part of #980 > --- > src/lib/core/coio.cc | 17 +++++++++++++++++ > 1 file changed, 17 insertions(+) > > diff --git a/src/lib/core/coio.cc b/src/lib/core/coio.cc > index e88d724d5..faa7e5bd5 100644 > --- a/src/lib/core/coio.cc > +++ b/src/lib/core/coio.cc > @@ -238,6 +238,17 @@ coio_connect_timeout(struct ev_io *coio, struct uri *uri, struct sockaddr *addr, > tnt_raise(SocketError, sio_socketname(coio->fd), "connection failed"); > } > > +/* Do not allow to reuse coio by different fiber. */ > +static inline void > +check_coio_in_use(struct ev_io *coio) > +{ > + if (ev_is_active(coio)) { > + errno = EINPROGRESS; > + tnt_raise(SocketError, sio_socketname(coio->fd), > + "already in use"); > + } > +} > + > /** > * Wait a client connection on a server socket until > * timedout. > @@ -249,6 +260,7 @@ coio_accept(struct ev_io *coio, struct sockaddr *addr, > ev_tstamp start, delay; > coio_timeout_init(&start, &delay, timeout); > > + check_coio_in_use(coio); > CoioGuard coio_guard(coio); > > while (true) { > @@ -302,6 +314,7 @@ coio_read_ahead_timeout(struct ev_io *coio, void *buf, size_t sz, > > ssize_t to_read = (ssize_t) sz; > > + check_coio_in_use(coio); > CoioGuard coio_guard(coio); > > while (true) { > @@ -399,6 +412,7 @@ coio_write_timeout(struct ev_io *coio, const void *buf, size_t sz, > ev_tstamp start, delay; > coio_timeout_init(&start, &delay, timeout); > > + check_coio_in_use(coio); > CoioGuard coio_guard(coio); > > while (true) { > @@ -461,6 +475,7 @@ coio_writev_timeout(struct ev_io *coio, struct iovec *iov, int iovcnt, > struct iovec *end = iov + iovcnt; > ev_tstamp start, delay; > coio_timeout_init(&start, &delay, timeout); > + check_coio_in_use(coio); > CoioGuard coio_guard(coio); > > /* Avoid a syscall in case of 0 iovcnt. */ > @@ -518,6 +533,7 @@ coio_sendto_timeout(struct ev_io *coio, const void *buf, size_t sz, int flags, > ev_tstamp start, delay; > coio_timeout_init(&start, &delay, timeout); > > + check_coio_in_use(coio); > CoioGuard coio_guard(coio); > > while (true) { > @@ -563,6 +579,7 @@ coio_recvfrom_timeout(struct ev_io *coio, void *buf, size_t sz, int flags, > ev_tstamp start, delay; > coio_timeout_init(&start, &delay, timeout); > > + check_coio_in_use(coio); > CoioGuard coio_guard(coio); > > while (true) { > -- > 2.25.0 -- Konstantin Osipov, Moscow, Russia
next prev parent reply other threads:[~2020-03-19 18:09 UTC|newest] Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top 2020-02-12 9:39 [Tarantool-patches] [PATCH v4 00/11] Replication from memory Georgy Kirichenko 2020-02-12 9:39 ` [Tarantool-patches] [PATCH v4 01/11] recovery: do not call recovery_stop_local inside recovery_delete Georgy Kirichenko 2020-03-19 7:55 ` Konstantin Osipov 2020-02-12 9:39 ` [Tarantool-patches] [PATCH v4 02/11] recovery: do not throw an error Georgy Kirichenko 2020-03-19 7:56 ` Konstantin Osipov 2020-02-12 9:39 ` [Tarantool-patches] [PATCH v4 03/11] coio: do not allow parallel usage of coio Georgy Kirichenko 2020-03-19 18:09 ` Konstantin Osipov [this message] 2020-02-12 9:39 ` [Tarantool-patches] [PATCH v4 04/11] coio: do not throw an error, minor refactoring Georgy Kirichenko 2020-03-23 6:59 ` Konstantin Osipov 2020-02-12 9:39 ` [Tarantool-patches] [PATCH v4 05/11] xstream: get rid of an exception Georgy Kirichenko 2020-02-12 9:39 ` [Tarantool-patches] [PATCH v4 06/11] wal: extract log write batch into a separate routine Georgy Kirichenko 2020-02-12 9:39 ` [Tarantool-patches] [PATCH v4 07/11] wal: matrix clock structure Georgy Kirichenko 2020-02-12 9:39 ` [Tarantool-patches] [PATCH v4 08/11] wal: track relay vclock and collect logs in wal thread Georgy Kirichenko 2020-02-12 9:39 ` [Tarantool-patches] [PATCH v4 09/11] wal: xrow memory buffer and cursor Georgy Kirichenko 2020-02-12 9:39 ` [Tarantool-patches] [PATCH v4 10/11] wal: use a xrow buffer object for entry encoding Georgy Kirichenko 2020-02-12 9:39 ` [Tarantool-patches] [PATCH v4 11/11] replication: use wal memory buffer to fetch rows Georgy Kirichenko
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=20200319180912.GA3662@atlas \ --to=kostja.osipov@gmail.com \ --cc=georgy@tarantool.org \ --cc=tarantool-patches@dev.tarantool.org \ --subject='Re: [Tarantool-patches] [PATCH v4 03/11] coio: do not allow parallel usage of coio' \ /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
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox