search for: _handle_single_connection

...004066ff in filter_free (b=0x203c330) at filters.c:77 #3 0x000000000040a6f4 in main (argc=11, argv=0x7ffc1f4486e8) at main.c:649 Thread 1 (Thread 0x7f1caaa5e700 (LWP 7226)): #0 0x000000000040732a in filter_finalize (b=0x203c330, conn=0x203d870) at filters.c:421 #1 0x0000000000404d07 in _handle_single_connection (sockin=6, sockout=6) at connections.c:239 #2 0x0000000000404d76 in handle_single_connection (sockin=6, sockout=6) at connections.c:258 #3 0x00000000004119f6 in start_thread (datav=0x203b450) at sockets.c:263 #4 0x00007f1cab09b5a2 in start_thread () from /lib64/libpthread.so.0 #5...

...ll be different). --- src/connections.c | 4 ++-- src/internal.h | 1 + src/plugins.c | 1 + 3 files changed, 4 insertions(+), 2 deletions(-) diff --git a/src/connections.c b/src/connections.c index 921a5b2..e1ac543 100644 --- a/src/connections.c +++ b/src/connections.c @@ -221,7 +221,7 @@ _handle_single_connection (int sockin, int sockout) if (backend->open (backend, conn, readonly) == -1) goto done; - threadlocal_set_name (backend->name (backend)); + threadlocal_set_name (backend->plugin_name (backend)); /* Handshake. */ if (negotiate_handshake (conn) == -1) @@ -253,7 +253,7 @@...

...pe[1] >= 0); if (write (conn->status_pipe[1], &c, 1) != 1 && errno != EAGAIN) - nbdkit_debug ("failed to notify pipe-to-self: %m"); + debug ("failed to notify pipe-to-self: %m"); } conn->status = value; } @@ -176,7 +176,7 @@ _handle_single_connection (int sockin, int sockout) debug ("handshake complete, processing requests with %d threads", nworkers); workers = calloc (nworkers, sizeof *workers); - if (!workers) { + if (unlikely (!workers)) { perror ("malloc"); goto done; } @@ -...

This fixes the long-standing crash on close when nbdkit exits. I did try first to fix threads so we're using a proper thread pool, but that's difficult to implement. So this does the minimal change needed to fix the crash instead. There are still two segfaults that happen during running the test suite. One is deliberately caused (tests/test-captive.sh). The other appears to be an

I'm fairly sure that these bugs which appear in the Python plugin: https://bugzilla.redhat.com/show_bug.cgi?id=1566516 https://bugzilla.redhat.com/show_bug.cgi?id=1566522 are really bugs in the SERIALIZE_ALL_REQUESTS thread model. See the first patch for the full explanation. The second patch is a fix for a race condition which is probably nudged into being by the first patch. Now this

I want to make my nbd forwarding plugin fully parallel - but to do that, I first need to make nbdkit itself fully parallel ;) With this series, I was finally able to demonstrate out-of-order responses when using qemu-io (which is great at sending back-to-back requests prior to waiting for responses) coupled with the nbd file plugin (which has a great feature of rdelay and wdelay, to make it

...++ b/src/Makefile.am @@ -40,6 +40,7 @@ nbdkit_SOURCES = \ crypto.c \ errors.c \ internal.h \ + locks.c \ main.c \ plugins.c \ protocol.h \ diff --git a/src/connections.c b/src/connections.c index 111a810..74bb8e4 100644 --- a/src/connections.c +++ b/src/connections.c @@ -211,7 +211,7 @@ _handle_single_connection (int sockin, int sockout) int nworkers = threads ? threads : DEFAULT_PARALLEL_REQUESTS; pthread_t *workers = NULL; - if (!plugin_is_parallel() || nworkers == 1) + if (plugin_thread_model () < NBDKIT_THREAD_MODEL_PARALLEL || nworkers == 1) nworkers = 0; conn = new_connection (so...

...++ b/src/Makefile.am @@ -40,6 +40,7 @@ nbdkit_SOURCES = \ crypto.c \ errors.c \ internal.h \ + locks.c \ main.c \ plugins.c \ protocol.h \ diff --git a/src/connections.c b/src/connections.c index 111a810..74bb8e4 100644 --- a/src/connections.c +++ b/src/connections.c @@ -211,7 +211,7 @@ _handle_single_connection (int sockin, int sockout) int nworkers = threads ? threads : DEFAULT_PARALLEL_REQUESTS; pthread_t *workers = NULL; - if (!plugin_is_parallel() || nworkers == 1) + if (plugin_thread_model () < NBDKIT_THREAD_MODEL_PARALLEL || nworkers == 1) nworkers = 0; conn = new_connection (so...

...nections.c @@ -59,6 +59,9 @@ /* Maximum length of any option data (bytes). */ #define MAX_OPTION_LENGTH 4096 +/* Default number of parallel requests. */ +#define DEFAULT_PARALLEL_REQUESTS 16 + /* Connection structure. */ struct connection { pthread_mutex_t request_lock; @@ -203,9 +206,11 @@ _handle_single_connection (int sockin, int sockout) { int r = -1; struct connection *conn; - int nworkers = 1; /* TODO default to 16 for parallel plugins, with command-line override */ + int nworkers = threads ? threads : DEFAULT_PARALLEL_REQUESTS; pthread_t *workers = NULL; + if (!plugin_is_parallel()) + n...

...2.7.so.1.0 > #9 0x00007ffff73f0af0 in py_close (handle=0x625338) at python.c:234 > #10 0x0000000000405803 in plugin_close (conn=conn@entry=0x7fffe8000910) at plugins.c:377 > #11 0x00000000004037ec in free_connection (conn=0x7fffe8000910) at connections.c:147 > #12 0x0000000000404476 in _handle_single_connection (sockout=<optimized out>, sockin=<optimized out>) at connections.c:99 > #13 handle_single_connection (sockin=<optimized out>, sockout=<optimized out>) at connections.c:109 > #14 0x0000000000405e73 in start_thread (datav=0x7fffffffdd40) at sockets.c:220 > #15 0x00007...

I've posted some of these patches or ideas before; but now I'm confident enough with the series that it should be ready to push; at any rate, I can now run test-socket-activation in a tight loop without triggering any crashes or hangs. With this in place, I'm going back to work on making the nbd forwarder wort with the parallel thread model. Eric Blake (8): sockets: Use

...-linux-gnu/libpython2.7.so.1.0 #9 0x00007ffff73f0af0 in py_close (handle=0x625338) at python.c:234 #10 0x0000000000405803 in plugin_close (conn=conn@entry=0x7fffe8000910) at plugins.c:377 #11 0x00000000004037ec in free_connection (conn=0x7fffe8000910) at connections.c:147 #12 0x0000000000404476 in _handle_single_connection (sockout=<optimized out>, sockin=<optimized out>) at connections.c:99 #13 handle_single_connection (sockin=<optimized out>, sockout=<optimized out>) at connections.c:109 #14 0x0000000000405e73 in start_thread (datav=0x7fffffffdd40) at sockets.c:220 #15 0x00007ffff79c0184 in...

...gt; > Thread 1 (Thread 0x7f1caaa5e700 (LWP 7226)): > #0 0x000000000040732a in filter_finalize (b=0x203c330, conn=0x203d870) > at filters.c:421 ...filter_finalize() time to finish its job? But I agree that solving the race is an independent patch. > #1 0x0000000000404d07 in _handle_single_connection (sockin=6, sockout=6) > at connections.c:239 > #2 0x0000000000404d76 in handle_single_connection (sockin=6, sockout=6) > at connections.c:258 > #3 0x00000000004119f6 in start_thread (datav=0x203b450) at sockets.c:263 > #4 0x00007f1cab09b5a2 in start_thread () from /li...

...00000000040685d in nbdkit_debug (fs=fs@entry=0x40ace6 "close") at errors.c:91 #4 0x00000000004077d9 in plugin_close (conn=conn@entry=0x7f679c000910) at plugins.c:386 #5 0x0000000000404de3 in free_connection (conn=0x7f679c000910) at connections.c:228 #6 0x00000000004059fd in _handle_single_connection (sockout=<optimized out>, sockin=<optimized out>) at connections.c:179 #7 handle_single_connection (sockin=<optimized out>, sockout=<optimized out>) at connections.c:189 #8 0x0000000000408023 in start_thread (datav=0x7fff1a4fbbe0) at sockets.c:262 #9 0x00007f67a8...

...h->can_write = 0; + if (h->can_write == -1) h->can_write = b->can_write (b, conn); - } return h->can_write; } diff --git a/server/connections.c b/server/connections.c index 7609e9a7..0c1f2413 100644 --- a/server/connections.c +++ b/server/connections.c @@ -149,7 +149,7 @@ _handle_single_connection (int sockin, int sockout) goto done; lock_request (conn); - r = backend->open (backend, conn, readonly); + r = backend_open (backend, conn, readonly); unlock_request (conn); if (r == -1) goto done; diff --git a/server/filters.c b/server/filters.c index 0e10816f..a8d7dd7c 10...

v1 -> v2: - Fixed everything mentioned in the review. Rich.

Somewhat invasive but mostly mechanical change to how plugins are called. This patch is in preparation for adding a second backend subtype for filters. Rich.

...repare (b, conn) == -1) { + backend_finalize (b, conn); + backend_close (b, conn); + return -1; + } + return 0; } int64_t diff --git a/server/connections.c b/server/connections.c index c6fa232f..c55d3816 100644 --- a/server/connections.c +++ b/server/connections.c @@ -216,10 +216,7 @@ _handle_single_connection (int sockin, int sockout) /* Finalize (for filters), called just before close. */ lock_request (conn); - if (backend) - r = backend_finalize (backend, conn); - else - r = 0; + r = backend_finalize (backend, conn); unlock_request (conn); if (r == -1) goto done; -- 2.21.0

...ugin.so #6 0x00007fc5d969833c in sh_close () from /usr/lib64/nbdkit/plugins/nbdkit-sh-plugin.so #7 0x0000561694bcf8e8 in plugin_close (b=<optimized out>, conn=0x56169df9afe0) at plugins.c:305 #8 0x0000561694bca649 in free_connection (conn=0x56169df9afe0) at connections.c:377 #9 _handle_single_connection (sockout=<optimized out>, sockin=<optimized out>) at connections.c:267 #10 handle_single_connection (sockin=<optimized out>, sockout=<optimized out>) at connections.c:277 #11 0x0000561694bc9583 in start_serving () at main.c:856 #12 main (argc=<optimized out>, a...

The concern is a client is blocked while processing a request. The nbdkit server design requires a thread per request being processed regardless of the number of connections or clients. We want to run 1000's of requests in parallel without needing a thread at nbdkit layer per request in flight. Our plugin layer is built around boost asio and a few threads in a worker pool running an io