search for: nbdkit_nanosleep

...--- server/public.c | 38 ++++++++++++++++++++++++-------------- 1 file changed, 24 insertions(+), 14 deletions(-) diff --git a/server/public.c b/server/public.c index a992df1..630de9b 100644 --- a/server/public.c +++ b/server/public.c @@ -304,6 +304,16 @@ nbdkit_realpath (const char *path) int nbdkit_nanosleep (unsigned sec, unsigned nsec) { + struct timespec ts; + + if (sec >= INT_MAX - nsec / 1000000000) { + nbdkit_error ("sleep request is too long"); + errno = EINVAL; + return -1; + } + ts.tv_sec = sec + nsec / 1000000000; + ts.tv_nsec = nsec % 1000000000; + #if defined HA...

...id processing our non-fatal signal in the same thread as any filter/plugin processing, or by paying attention to EINTR results and restarting a sleep. And rather than making each sleeping client reimplement that logic, we might as well stick it in on our common helper. Thus, this patch introduces nbdkit_nanosleep, and implements it under the hood as a ppoll that redirects signals but wakes up as needed. Including <time.h> for struct timespec in nbdkit-common.h is difficult (C++ is not required to provide it), so we just take the two arguments in pieces; we can rely on int being 32-bits, so don't n...

On 8/3/19 11:01 AM, Eric Blake wrote: > There are a couple of problems with filters trying to sleep. First, > when it is time to shut down nbdkit, we wait until all pending > transactions have had a chance to wind down. But consider what > happens if one or more of those pending transactions are blocked in a > sleep. POSIX says nanosleep is interrupted with EINTR if that thread

On 8/28/19 11:14 AM, Richard W.M. Jones wrote: > Rather than failing to compile on platforms which lack POLLRDHUP such > as FreeBSD, simply fallback to the old method of sleeping. > > This leaves the porting suggestions as a comment in case someone wants > to implement a better solution for particular platforms. > --- > server/public.c | 38

...oint where the main thread has tried to unload the plugin, which is never a nice thing. We may still have more work ahead of us to ensure that we don't unload an in-use plugin. Eric Blake (3): server: Add threadlocal_get_conn server: Add pipe for tracking disconnects server: Add and use nbdkit_nanosleep docs/nbdkit-plugin.pod | 28 +++++++++++++++++ configure.ac | 1 + include/nbdkit-common.h | 1 + server/internal.h | 3 ++ filters/delay/delay.c | 14 ++------- filters/rate/rate.c | 10 +++---- server/connections.c | 66 ++++++++++++++++++++++++++++++++++++++++- se...

...c */ extern struct backend *filter_register (struct backend *next, size_t index, @@ -515,6 +517,11 @@ extern void *threadlocal_buffer (size_t size); extern void threadlocal_set_conn (struct connection *conn); extern struct connection *threadlocal_get_conn (void); +/* public.c */ +extern int do_nbdkit_nanosleep (unsigned sec, unsigned nsec); +extern const char *do_nbdkit_export_name (void); +extern int do_nbdkit_peer_name (struct sockaddr *addr, socklen_t *addrlen); + /* Macro which sets local variable struct connection *conn from * thread-local storage, asserting that it is non-NULL. If you want *...

The test relies on the timeout program. Also, since the nbdkit_nanosleep code relies on the Linux extension POLLRDHUP to detect early client closure, we may have to relax that part of the test when porting to platforms that lack ppoll/POLLRDHUP. (That is, while we should still be able to let a signal to the server shut down nbdkit quickly, it's harder to let a clie...

...n *threadlocal_get_conn (void); +#define GET_CONN (threadlocal_get_conn ()) /* Declare program_name. */ #if HAVE_DECL_PROGRAM_INVOCATION_SHORT_NAME == 1 diff --git a/server/public.c b/server/public.c index 418945fe..8fa7e21b 100644 --- a/server/public.c +++ b/server/public.c @@ -533,7 +533,7 @@ nbdkit_nanosleep (unsigned sec, unsigned nsec) * NBD_CMD_DISC or a problem with the connection * - the input socket detects POLLRDHUP/POLLHUP/POLLERR */ - struct connection *conn = threadlocal_get_conn (); + struct connection *conn = GET_CONN; struct pollfd fds[] = { [0].fd = quit_fd, [...

On 3/24/20 3:12 PM, Eric Blake wrote: >> (For non-mingw platforms) this breaks the source API promises rather >> seriously, so if I understand your proposal correctly I don't think >> this is a good idea.  It's possibly something we can consider for >> internal plugins, or for the V3 API. > > How does it break API to request that someone link against a

...I had a look into how we might implement libnbdkit.so. Some functions > are obviously self-contained (eg. nbdkit_parse_*, nbdkit_realpath, > nbdkit_debug, nbdkit_error, nbdkit_*extents). > > Unfortunately some functions depend themselves on internals > of the server: > > * nbdkit_nanosleep, nbdkit_export_name, nbdkit_peer_name call > threadlocal_get_conn > * nbdkit_set_error calls threadlocal_set_error > * nbdkit_shutdown must set the quit global (or call a server function) Yeah, there's some awkward dependencies to figure out. It's obvious the library has...

...ouldn't be a source API break. I had a look into how we might implement libnbdkit.so. Some functions are obviously self-contained (eg. nbdkit_parse_*, nbdkit_realpath, nbdkit_debug, nbdkit_error, nbdkit_*extents). Unfortunately some functions depend themselves on internals of the server: * nbdkit_nanosleep, nbdkit_export_name, nbdkit_peer_name call threadlocal_get_conn * nbdkit_set_error calls threadlocal_set_error * nbdkit_shutdown must set the quit global (or call a server function) I guess we can deal with the first ones by moving threadlocal.c into the same library, although it's a bit...

...#include <stdarg.h> #include <stdint.h> #include <errno.h> +#include <sys/socket.h> #include <nbdkit-version.h> @@ -87,6 +88,7 @@ extern int nbdkit_read_password (const char *value, char **password); extern char *nbdkit_realpath (const char *path); extern int nbdkit_nanosleep (unsigned sec, unsigned nsec); extern const char *nbdkit_export_name (void); +extern int nbdkit_peer_name (struct sockaddr *addr, socklen_t *addrlen); struct nbdkit_extents; extern int nbdkit_add_extent (struct nbdkit_extents *, diff --git a/server/nbdkit.syms b/server/nbdkit.syms index 1fb131...

This happened on s390 in Koji, once. The key lines from the log are: + start_t=0 nbdkit: sh[1]: debug: retry 1: waiting 1 seconds before retrying nbdkit: sh[1]: debug: retry 1: waiting 1 seconds before retrying + end_t=1 Somehow nbdkit did wait 2 seconds (or at least, nbdkit_nanosleep (1, 0) was called twice by the retry filter). But in the bash script, time (as measured by $SECONDS) advanced from 0 to 1. I'm not sure I can explain that ... It should be possible to make the test more robust by increasing the retry time and subtracting a bit of slop in the test of end_t -...

...64_t *r); -extern int nbdkit_parse_uint64_t (const char *what, const char *str, - uint64_t *r); -extern int nbdkit_stdio_safe (void); -extern int nbdkit_read_password (const char *value, char **password); -extern char *nbdkit_realpath (const char *path); -extern int nbdkit_nanosleep (unsigned sec, unsigned nsec); -extern int nbdkit_peer_name (struct sockaddr *addr, socklen_t *addrlen); -extern void nbdkit_shutdown (void); +extern NBDKIT_DLLEXPORT char *nbdkit_absolute_path (const char *path); +extern NBDKIT_DLLEXPORT int64_t nbdkit_parse_size (const char *str); +extern NBDKIT_...

...lay-shutdown.sh @@ -37,7 +37,7 @@ requires qemu-io --version requires timeout --version sock=`mktemp -u` -files="shutdown.pid $sock" +files="delay-shutdown.pid $sock" cleanup_fn rm -f $files fail=0 @@ -45,7 +45,7 @@ fail=0 # This tests that the delay filter's use of nbdkit_nanosleep is able to # react to both connection death and server shutdown without finishing # the entire delay duration. -start_nbdkit -P shutdown.pid -U $sock --filter=noparallel --filter=delay \ +start_nbdkit -P delay-shutdown.pid -U $sock --filter=noparallel --filter=delay \ null 1M serialize=conne...

This is the previous 5/9 patch posted earlier today, split into reviewable chunks. This passes bisection with -x 'make && make check', but I didn't work very hard on the commit messages, so I refer you back to the original patch to explain how it works: https://www.redhat.com/archives/libguestfs/2020-March/msg00248.html Rich.

...ht implement libnbdkit.so. Some functions > >are obviously self-contained (eg. nbdkit_parse_*, nbdkit_realpath, > >nbdkit_debug, nbdkit_error, nbdkit_*extents). > > > >Unfortunately some functions depend themselves on internals > >of the server: > > > > * nbdkit_nanosleep, nbdkit_export_name, nbdkit_peer_name call > > threadlocal_get_conn > > * nbdkit_set_error calls threadlocal_set_error > > * nbdkit_shutdown must set the quit global (or call a server function) > > Yeah, there's some awkward dependencies to figure out. It's obv...

...; + functions->parse_uint32_t = nbdkit_parse_uint32_t; + functions->parse_int64_t = nbdkit_parse_int64_t; + functions->parse_uint64_t = nbdkit_parse_uint64_t; + functions->read_password = nbdkit_read_password; + functions->realpath = nbdkit_realpath; + functions->nanosleep = nbdkit_nanosleep; + functions->export_name = nbdkit_export_name; + functions->peer_name = dummy_peer_name; + functions->shutdown = nbdkit_shutdown; + functions->add_extent = nbdkit_add_extent; + functions->extents_new = nbdkit_extents_new; + functions->extents_free = nbdkit_extents_free; +...

...d on s390 in Koji, once. The key lines from the > log are: > > + start_t=0 > nbdkit: sh[1]: debug: retry 1: waiting 1 seconds before retrying > nbdkit: sh[1]: debug: retry 1: waiting 1 seconds before retrying > + end_t=1 > > Somehow nbdkit did wait 2 seconds (or at least, nbdkit_nanosleep (1, 0) > was called twice by the retry filter). But in the bash script, time > (as measured by $SECONDS) advanced from 0 to 1. FWIW $SECONDS is calculated by bash in the obvious way. It saves time(2) when the shell starts, and then when you ask for $SECONDS it returns the current time(2) -...

The previous incomplete patch was here: https://www.redhat.com/archives/libguestfs/2019-September/msg00049.html based on earlier discussion here: https://www.redhat.com/archives/libguestfs/2019-September/msg00047.html In v2: - The previous patch was incomplete. This version completes it by adding tests and extending nbdkit-sh-plugin. - nbdkit_export_name now returns NULL for error,