similar to: PATCH: atomicio doesn't poll()

http://bugzilla.mindrot.org/show_bug.cgi?id=1366 Summary: atomicio doesn't poll() Product: Portable OpenSSH Version: 4.7p1 Platform: All OS/Version: All Status: NEW Severity: normal Priority: P2 Component: Miscellaneous AssignedTo: bitbucket at mindrot.org ReportedBy: ben at psc.edu

FYI in order to get 2.1.1p2 to work on my HP-UX 11.0 systems I had to patch atomicio.c for EWOULDBLOCK (HP read() does not give the POSIX return code). The new atomicio() is a clean fix for this timing problem; all it needs now is this one little tweak. Also had the "General Commercial Security" error (PAM_TERM_ERROR from pam_acct_mgmt()) which I have very crudely addressed for now by

Hi, At the n2k10 OpenBSD network hackathon, I finally got some time to clean up and rewrite the ssh(1) client multiplexing code. The attached diffs (one for portable OpenSSH, one for OpenBSD) are the result, and they need some testing. The revised multiplexing code uses a better protocol between the master and slave processes and I even bothered to write it up :) It tracks the control sockets

https://bugzilla.mindrot.org/show_bug.cgi?id=1467 Summary: SFTP stalls Classification: Unclassified Product: Portable OpenSSH Version: 5.0p1 Platform: Other OS/Version: Other Status: NEW Severity: normal Priority: P2 Component: ssh AssignedTo: bitbucket at mindrot.org ReportedBy: sconeu

On Wed, Jul 15, 2020 at 03:32:55PM +0200, mwilck at suse.com wrote: > From: Martin Wilck <mwilck at suse.com> > > If a program opens /dev/hwrng with O_NONBLOCK and uses poll() and > non-blocking read() to retrieve random data, it ends up in a tight > loop with poll() always returning POLLIN and read() returning EAGAIN. > This repeats forever until some process makes a

On 11/08/2020 12:37, Philippe Mathieu-Daud? wrote: > You Cc'ed qemu-devel, so Cc'ing the virtio-rng maintainers. > > On 7/15/20 3:32 PM, mwilck at suse.com wrote: >> From: Martin Wilck <mwilck at suse.com> >> >> If a program opens /dev/hwrng with O_NONBLOCK and uses poll() and >> non-blocking read() to retrieve random data, it ends up in a tight

On 28/08/2020 23:34, Martin Wilck wrote: > On Wed, 2020-08-26 at 08:26 -0400, Michael S. Tsirkin wrote: >> On Tue, Aug 11, 2020 at 04:42:32PM +0200, Laurent Vivier wrote: >>> On 11/08/2020 16:28, mwilck at suse.com wrote: >>>> From: Martin Wilck <mwilck at suse.com> >>>> >>>> If a program opens /dev/hwrng with O_NONBLOCK and uses poll()

> So the make process gets an EAGAIN on the write syscall and doesn't > retry? That sounds like a bug in whatever make you're using, since > that could potentially occur in other circumstances too. What other circumstances? EAGAIN means that something put the device into non-blocking mode, and normally, that should only happen if the program calling write had itself previously

On 11/08/2020 16:28, mwilck at suse.com wrote: > From: Martin Wilck <mwilck at suse.com> > > If a program opens /dev/hwrng with O_NONBLOCK and uses poll() and > non-blocking read() to retrieve random data, it ends up in a tight > loop with poll() always returning POLLIN and read() returning EAGAIN. > This repeats forever until some process makes a blocking read() call.

On 11/08/2020 16:28, mwilck at suse.com wrote: > From: Martin Wilck <mwilck at suse.com> > > If a program opens /dev/hwrng with O_NONBLOCK and uses poll() and > non-blocking read() to retrieve random data, it ends up in a tight > loop with poll() always returning POLLIN and read() returning EAGAIN. > This repeats forever until some process makes a blocking read() call.

On Thu, 26 Feb 2015, Damien Miller wrote: > > Yes I saw that later. > > > > The testsuite build fails on Solaris 2.6 thusly: > [...] > > regress/netcat.c:1037: error: 'struct msghdr' has no member named > > 'msg_control' > > ah, looks like we need to copy some bits from monitor_fdpass.c Perhaps like this: diff --git regress/netcat.c

Hi, This diff (for portable) makes the chrooted preauth privsep process log via the monitor using a shared socketpair. It removes the need for /dev/log inside /var/empty and makes mandatory sandboxing of the privsep child easier down the road (no more socket() syscall required). Please test. -d Index: log.c =================================================================== RCS file:

https://bugzilla.mindrot.org/show_bug.cgi?id=1386 Summary: OpenSSH 4.7p1 compile error in atomicio.c under Tru64 4.0f Classification: Unclassified Product: Portable OpenSSH Version: 4.7p1 Platform: Alpha OS/Version: Tru64 Status: NEW Severity: major Priority: P2 Component: Build

tridge and Wayne in particular: I checked in this patch, which is meant to consolidate the ones from both of you for handling EOF in a modules list. The idea is that we need to handle servers that just close the socket rather than sending a nice ending token, but we want to keep EOF detection on in general. (The IO code is such a mess!) -- Martin Index: clientserver.c

Previously we performed a single call to recv(2) or send(2) (or the GnuTLS equivalents), and even if more data/space was immediately available to receive/send we would return to poll. Instead of this, loop until the socket returns EAGAIN. --- generator/states.c | 91 ++++++++++++++++++++++------------------------ 1 file changed, 43 insertions(+), 48 deletions(-) diff --git a/generator/states.c

I have a fix for the sendfile.c code for HP-UX - length not properly specified prior to call. I've also added code to support the AIX send_file() function and configure.in code to check for AIX specific send_file API and sets AIX_SENDFILE_API if successful. Could you consider this patch for 3.0.11? I also sent a web/cgi.c patch for "struct var" last week with respect to compile

When copying from a remote host to a local filesystem with a strange block size, allocbuf() in scp.c seems to calculate an incorrect buffer size, causing the copy loop in sink() to write past the end of the buffer. For example, with smbfs, the optimal block size is negotiated when the client connects to the server, and is rarely a power of two. In my case it is 64560. This loop in sink() keeps

Hello, I ran into a few problems when trying to build openssh 5.2p1 on IRIX 5.3. First one is new to 5.2p1: cc -I. -I. -I/usr/tgcware/include/openssl -I/usr/tgcware/include -DSSHDIR=\"/usr/tgcware/etc/ssh\" -D_PATH_SSH_PROGRAM=\"/usr/tgcware/bin/ssh\" -D_PATH_SSH_ASKPASS_DEFAULT=\"/usr/tgcware/libexec/ssh-askpass\"

I noted a possible problem restoring a machine. In xc_domain_restore (xc_domain_restore.c) if it''s not the last checkpoint we set O_NONBLOCK flag (search for fcntl) that we can call pagebuf_get or just load other pages (see following "goto loadpages;" line). Now we could ending up calling xc_tmem_restore/xc_tmem_restore_extra (xc_tmem.c) which call read_extract (xc_private.c)

First the bug: I've found a timing problem in 2.1.1p1 at the point where the client version string is read, a core dump with a "Did not receive ident string..." error. This problem does not appear to have been mentioned yet in the list archive. This bug was noted on HP-UX 11.0 but could be a problem on other Unices as well. My Q&D fix (patch below) was to spin on EWOULDBLOCK