openafs.git
3 months agoMake OpenAFS 1.8.6pre1 47/14047/4 openafs-stable-1_8_6pre1
Stephan Wiesand [Sat, 25 Jan 2020 19:32:26 +0000]
Make OpenAFS 1.8.6pre1

Update version strings for the first 1.8.6 prerelease.

Change-Id: I8e63f8532970a8b0f8d3a39d93d9ae07e83f9df7
Reviewed-on: https://gerrit.openafs.org/14047
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Yadavendra Yadav <yadayada@in.ibm.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoUpdate NEWS for OpenAFS 1.8.6pre1 70/14070/3
Stephan Wiesand [Mon, 17 Feb 2020 19:49:02 +0000]
Update NEWS for OpenAFS 1.8.6pre1

Release notes for the first 1.8.6 prerelease

Change-Id: I22912d4526d079c0819e156a4e6fe1d38fdeee35
Reviewed-on: https://gerrit.openafs.org/14070
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Yadavendra Yadav <yadayada@in.ibm.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoLINUX 5.6: use struct proc_ops for proc_create 69/14069/2
Cheyenne Wills [Fri, 14 Feb 2020 18:50:03 +0000]
LINUX 5.6: use struct proc_ops for proc_create

The Linux commit d56c0d45f0e27f814e87a1676b6bdccccbc252e9
(proc: decouple proc from VFS with "struct proc_ops") was merged into
Linux 5.6rc1.  The commit replaces the 'file_operations' parameter for
proc_create with a new structure 'proc_ops'.

Conditionally initialize and use proc_ops structures instead of
file_operations structures for calls to proc_create.

Notes:
 * proc_ops.proc_ioctl is equivalent to file_operations.unlocked_ioctl
 * The macros HAVE_UNLOCKED_IOCTL and HAVE_COMPAT_IOCTL are both
   hardcoded to 1 in linux's fs.h
 * proc_ops.compat_ioctl is conditional on Linux's CONFIG_COMPAT macro
   which is a separate test from the HAVE_COMPAT_IOCTL macro

Reviewed-on: https://gerrit.openafs.org/14063
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 1626986bd6d70c526376cf7cedfd3ebbf6d3588a)

Change-Id: Icaab45f4542131e636f2c60e3efce86c8afc57be
Reviewed-on: https://gerrit.openafs.org/14069
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Yadavendra Yadav <yadayada@in.ibm.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agomacos: add anchors to synthetic.conf grep pattern 68/14068/2
Marcio Barbosa [Fri, 7 Feb 2020 17:58:56 +0000]
macos: add anchors to synthetic.conf grep pattern

The grep pattern that checks if /etc/synthetic.conf already has an entry
for afs is intended to check if this file holds a single column entry
named afs. Unfortunately, the current version does not completely
enforce this restriction. To fix this problem, add anchors to the grep
pattern in question.

Reviewed-on: https://gerrit.openafs.org/14062
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Yadavendra Yadav <yadayada@in.ibm.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 6d6a28720f4eae4652f2628fdfcc30983916f39d)

Change-Id: Iea837157a9eb5c066d577c705c445e10e244757d
Reviewed-on: https://gerrit.openafs.org/14068
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agomacos: add entry for afs into synthetic.conf 37/14037/3
Marcio Barbosa [Sun, 22 Dec 2019 03:56:41 +0000]
macos: add entry for afs into synthetic.conf

The root mount point is read-only as of macOS 10.15. As a result, /afs
cannot be created at this location. To workaround this restriction,
macOS 10.15 provides an alternative way to create mount points at the
root. To make it possible, an entry for the mount point in question must
be added to /etc/synthetic.conf. The synthetic entities described in
this file are not physically present on the disk. Instead, they are
synthesized by the kernel during system boot.

This commit adds an entry for afs into the file mentioned above. Knowing
that this change only takes effect after reboot, also provide directions
to the user during the installation process.

Reviewed-on: https://gerrit.openafs.org/13928
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Yadavendra Yadav <yadayada@in.ibm.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit daf6616aab6732d6b417c15f6f401731ef8e44b5)

Change-Id: If990608d968061ac8ab0391dbd83d1c6a87d32a6
Reviewed-on: https://gerrit.openafs.org/14037
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agomacos: add script to notarize OpenAFS 36/14036/3
Marcio Barbosa [Sun, 22 Dec 2019 03:11:57 +0000]
macos: add script to notarize OpenAFS

In order to integrate the notarization process into our existing build
scripts, this patch introduces a script to automatically notarize the
OpenAFS package.

Reviewed-on: https://gerrit.openafs.org/13671
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 0563642cc1cb750c69a6471005adf36fabb2b7e3)

Change-Id: I50265b3305eb12db45371da1bf1982a6722c0018
Reviewed-on: https://gerrit.openafs.org/14036
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agomacos: prepare for notarization 35/14035/3
Marcio Barbosa [Tue, 26 Nov 2019 19:41:36 +0000]
macos: prepare for notarization

With the public release of macOS 10.14.5, all new and updated kernel
extensions must be notarized by Apple. To be taken into consideration,
all executables must be signed and the Hardened Runtime capability must
be enabled.

This patch adds the missing prerequisites mentioned above.

Reviewed-on: https://gerrit.openafs.org/13670
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 63fd13bf9e6af21136007c9980816875ebea5f7c)

Change-Id: If0c27732f667945f430fd2c5698e8f58a84e3bde
Reviewed-on: https://gerrit.openafs.org/14035
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agomacos: packaging support for MacOS X 10.15 34/14034/3
Marcio Barbosa [Fri, 28 Jun 2019 03:40:55 +0000]
macos: packaging support for MacOS X 10.15

This commit introduces the new set of changes / files required to
successfully create the dmg installer on OS X 10.15 "Catalina".

Reviewed-on: https://gerrit.openafs.org/13669
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit c7864b73603842b8beaee03fcbb2426890205410)

Change-Id: I0b6b0616f2f2413a466c60986957e6bc3e21cbb0
Reviewed-on: https://gerrit.openafs.org/14034
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Yadavendra Yadav <yadayada@in.ibm.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agomacos: add support for MacOS 10.15 33/14033/3
Marcio Barbosa [Mon, 18 Nov 2019 14:34:08 +0000]
macos: add support for MacOS 10.15

This commit introduces the new set of changes / files required to
successfully build the OpenAFS source code on OS X 10.15 "Catalina".

Reviewed-on: https://gerrit.openafs.org/13668
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 93815caabc92acc6edc62b72805b44d2e46748cf)

Change-Id: Ia1fb98dd59d7b0ddad9c16c04b823623e07dd498
Reviewed-on: https://gerrit.openafs.org/14033
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agomacos: upgrade *.xib files 32/14032/3
Marcio Barbosa [Fri, 13 Dec 2019 03:03:04 +0000]
macos: upgrade *.xib files

According to Xcode 11, the *.xib files updated by this commit use an
older format that is potentially insecure when decoded. To fix this
problem, Xcode automatically upgraded these files to the modern format.
These changes are required to build OpenAFS on Catalina (Xcode 11).

Reviewed-on: https://gerrit.openafs.org/13935
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit d4302d42149988fa6d04d626967063dfa916c9fd)

Change-Id: I1e29493a8431d4ad13ff36762f6112dd5309573c
Reviewed-on: https://gerrit.openafs.org/14032
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agomacos: tell the compiler the system include path 31/14031/3
Marcio Barbosa [Fri, 8 Nov 2019 02:56:13 +0000]
macos: tell the compiler the system include path

In order to support multiple SDKs, macOS Catalina no longer has the
/usr/include directory. As a result, the compiler needs to know where
these headers can be found. To successfully build OpenAFS on OSX 10.15,
set KROOT so the compiler knows the correct location of these headers.

Reviewed-on: https://gerrit.openafs.org/13936
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 677b038814817defec9421e698ce67b44a7fd7d1)

Change-Id: I2043c2bc6e745ca55faf68b77d791168bc57bb1d
Reviewed-on: https://gerrit.openafs.org/14031
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoviced: add opt to allow admin writes on RO servers 19/14019/3
Marcio Barbosa [Thu, 14 Nov 2019 20:29:56 +0000]
viced: add opt to allow admin writes on RO servers

Add the new option -admin-write to allow write requests from superusers
on file servers running in readonly mode (-readonly). This lets sites
run fileservers in readonly mode for normal users, but allows members of
the system:administrators group to modify content.

Reviewed-on: https://gerrit.openafs.org/13707
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit f5f8b9336919debc5c26c429b12a14b65e0b697c)

Change-Id: Ia627b8c99767a875c1e8d1c69dcb45118df36937
Reviewed-on: https://gerrit.openafs.org/14019
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoviced: prevent writes on readonly fileservers 18/14018/3
Marcio Barbosa [Thu, 14 Nov 2019 04:15:47 +0000]
viced: prevent writes on readonly fileservers

Currently, a fileserver can be initialized as readonly. In this mode,
writes on this server should not be allowed. Unfortunately, updates on
files stored by readonly fileservers are not completely prevented. In
some situations, the check for RO server is omitted (e.g. if the user is
the owner of the file to be updated). In other situations, the same
check is redundant.

To fix these problems, consolidate this check in one place.

Reviewed-on: https://gerrit.openafs.org/13934
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 0593017177edd5b3bc6609d9dfcce55f15bba3e9)

Change-Id: I42034928d1f5e9342029121613ac8d716818c3ae
Reviewed-on: https://gerrit.openafs.org/14018
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoafs: Ensure CDirty is set during afs_write loop 51/13951/3
Andrew Deason [Mon, 18 Nov 2019 02:58:15 +0000]
afs: Ensure CDirty is set during afs_write loop

Currently, in afs_write(), we set CDirty on the given vcache, and then
write the given data into various dcaches. When writing to a dcache,
we call afs_DoPartialWrite, which may cause us to flush the dirty data
to the fileserver and clear the CDirty bit.

If we were given more than 1 chunk of data to write, we will then go
through another iteration of the loop, writing more dirty data into
dcaches, but CDirty will not be set. This can cause issues with, for
example, afs_SimpleVStat() or afs_ProcessFS(), which use CDirty to
determine whether or not to merge in FetchStatus info from the
fileserver into our local cache. This can cause our local cache to
incorrectly reflect the state of the file on the fileserver, instead
of the state of the locally-modified file in our cache.

A more detailed example is as follows. Consider a small C program that
copies a file, fchmod()ing the destination before closing it:

    void
    do_copy(char *src_name, char *dest_name)
    {
        /* error checking elided */
        src_fd = open(src_name, O_RDONLY);
        dest_fd = open(dest_name, O_WRONLY|O_CREAT|O_TRUNC, 0755);
        fstat(src_fd, &st);
        src_buf = mmap(NULL, st.st_size, PROT_READ, MAP_SHARED, src_fd, 0);
        write(dest_fd, src_buf, st.st_size);
        munmap(src_buf, st.st_size);
        close(src_fd);
        fchmod(dest_fd, 0100644);
        close(dest_fd);
    }

Currently, on FBSD, using this to copy a 7862648-byte file, using a
smallish cache (10000 blocks) will cause the destination to appear to
be truncated, because avc->f.m.Length will be incorrect, even though
all of the relevant data was written to the fileserver.

On most other platforms such as SOLARIS and LINUX, this is not a
problem, since currently they only write one page of data at a time to
afs_write(), and so they never hit multiple iterations of the while()
loop inside afs_write().

To fix this, just set CDirty on every iteration of the while() loop in
afs_write(). In general, we need to set CDirty after calling
afs_DoPartialStore() anywhere if the caller continues to write more
data. But all callers already do this, except for this one instance in
afs_write().

Thanks to tcreech@tcreech.com for helping find occurrences of the
relevant issue.

FIXES 135041

Reviewed-on: https://gerrit.openafs.org/13948
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 9d0854547522f7b2fb1bb7aa876fe9f901674747)

Change-Id: Ie86313e9b9750bc6724bb6e18b7df8e010810023
Reviewed-on: https://gerrit.openafs.org/13951
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoafs: Avoid -1 error for vreadUIO/vwriteUIO 38/13938/3
Andrew Deason [Tue, 5 Nov 2019 02:03:43 +0000]
afs: Avoid -1 error for vreadUIO/vwriteUIO

Commit c6b61a45 (afs: Verify osi_UFSOpen worked) added various checks
to return an error if a given osi_UFSOpen failed. However, two of
these checks (in afs_UFSReadUIO and afs_UFSWriteUIO) result in us
returning -1 on error, in functions that otherwise return errno codes
(e.g. ENOSPC). An error code of -1 might get interpreted as
RX_CALL_DEAD, which would be rather confusing, so use EIO as a generic
error instead.

Reviewed-on: https://gerrit.openafs.org/13931
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 360b9d5d71fb1de142ae4efd4660732476855a3f)

Change-Id: I4c6773affe02cc7a3ca01cf25bea21c960d98e87
Reviewed-on: https://gerrit.openafs.org/13938
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agovos: fix name availability check in vos rename 55/14055/2
Michael Meffie [Mon, 22 Jul 2019 19:20:24 +0000]
vos: fix name availability check in vos rename

The UV_RenameVolume() function first updates the volume name in the
VLDB, then read-write volume header and backup volume header, and
finally all of the read-only volume headers. If this function is
interrupted or a remote site is not reachable, the names in some of the
volume headers will be out of sync with name in the VLDB entry.

The implementation of UV_RenameVolume() is idempotent, so can be safely
called with the same name as in the volume's VLDB entry. This could be
used to bring all the names in the volume headers in sync with the name
in the VLDB.

Unfortunately, due to the check of the -newname parameter, vos
rename will not invoke UV_RenameVolume() when the name in the VLDB has
already been changed.  The vos rename command attempts to verify the
desired name (-newname) is available before invoking UV_RenameVolume()
by simply checking if a VLDB entry exists with that name, and
incorrectly assumes when a VLDB entry exists with that name it is an
entry for a different volume.

Change the -newname check to allow vos rename to proceed when name has
already been set in the VLDB entry of the volume being renamed. This
allows admins to run vos rename command to complete a previously
incomplete rename operation and bring the names in the volume headers in
sync with the name in the VLDB entry.

Note: Before this commit, administrators could workaround this vos
rename limitation by renaming the volume twice, first to an unused
volume name, then to the actual desired volume name.

Remove the useless checks of the code1 return code after exit in
the RenameVolume() function. These checks for code1 are never performed
since the function exits early when the first VLDB_GetEntryByName()
fails for any reason.

Update the vos rename man page to show vos rename can be used to fix
previously interrupted/failed rename. Also document the -oldname
parameter accepts a numeric volume id to specify the volume to be
renamed.

Reviewed-on: https://gerrit.openafs.org/13720
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 9238b1eb9ef02889855eaade76e5b7962e5f2f28)

Change-Id: I8b03e4211c5d306f55779130c8461b14bc4913f0
Reviewed-on: https://gerrit.openafs.org/14055
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoRedHat: Update makesrpm.pl to use @PACKAGE_VERSION@ instead of @VERSION@ 50/14050/2
Cheyenne Wills [Fri, 31 Jan 2020 23:14:40 +0000]
RedHat: Update makesrpm.pl to use @PACKAGE_VERSION@ instead of @VERSION@

Commit 2f2c2ce62aa17ecac3651d64c1168af926f7458b
'Remove automake autoconf vars' replaced the automake variable @VERSION@
with the autoconf variable @PACKAGE_VERSION@. (Gerrit #13357)

The RedHat openafs.spec.in is not processed using autoconf, but
by 'makesrpm.pl', which was not updated to use @PACKAGE_VERSION@.

Update makesprm.pl to use @PACKAGE_VERSION@ instead of @VERSION@

Reviewed-on: https://gerrit.openafs.org/13887
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit b03f3e6101ff21a6f148c555c213c47678482a7b)

Change-Id: I6bc27474b1b8dfa8b63806a4e0e996a00dd302e4
Reviewed-on: https://gerrit.openafs.org/14050
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoubik: improve logging for database synchonizations 08/13908/3
Mark Vitale [Fri, 4 May 2018 21:32:51 +0000]
ubik: improve logging for database synchonizations

As an aid for debugging database synchronization issues, ensure that the
logging is consistent and unambiguous for both the client and server
sides of DISK_GetFile and DISK_SendFile.  Add new error messages as
required.

In addition, rework the "recovery sending version to <IP>" message in
urecovery_Interact.  This message is misleading because the new version
database is only sent to a DB server if its version is not up to date.
Instead, move this message into the version check block immediately
below it.  Also reword it for clarity and promote its log level from 5
to 0.  Finally, remove the now-superfluous "recovery stating local
database" log message.

Reviewed-on: https://gerrit.openafs.org/13079
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 0e1c042615d1aeb919a22568cdd2b2ea42c677ba)

Change-Id: I26e876e5bcd5adc004b985ea8c3f716cb6a72b5d
Reviewed-on: https://gerrit.openafs.org/13908
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoubik: urecovery_AbortAll diagnostic msgs 07/13907/3
Mark Vitale [Fri, 17 Mar 2017 22:12:23 +0000]
ubik: urecovery_AbortAll diagnostic msgs

As a troubleshooting aid for developers, add a few counters and a log
msg so we know when transactions are being aborted (if any) by
urecovery_AbortAll.

Reviewed-on: https://gerrit.openafs.org/12618
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
(cherry picked from commit eac22d3e46c72c0e2b82f35c5187d50b6fa136a2)

Change-Id: Ia91bc1c5f041eccc9b974d4b195fed1a889252e7
Reviewed-on: https://gerrit.openafs.org/13907
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoubik: log important messages at default log level 06/13906/3
Mark Vitale [Tue, 9 May 2017 01:11:27 +0000]
ubik: log important messages at default log level

Many important ubik messages (e.g., errors, warnings, sync state
changes) are logged at log level 5 (-d 5) or higher.  Many sites are
reluctant to run ubik servers at a logging level higher than the default
due to the large number of extremely noisy informational messages at log
level 5.  Therefore, many important log messages are never seen.

Instead, issue critical errors, warnings, and other important messages
at log level 0 so that they are always seen, even at the default logging
level.

In addition, disambiguate the two "I am no longer sync-site" messages by
adding a unique reason text to each.

Reviewed-on: https://gerrit.openafs.org/12617
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
(cherry picked from commit 8b0e312d043d435f0e55c6dc14f5446ffedc7ce4)

Change-Id: I87425e78fb4f7fb1aa393b2f5b81ab34a71a38c4
Reviewed-on: https://gerrit.openafs.org/13906
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoubik: Save errno before logging 98/13898/5
Andrew Deason [Fri, 27 Jul 2018 18:36:15 +0000]
ubik: Save errno before logging

The value of errno can change after a syscall, and ViceLog may issue
syscalls (such as write()). So, make sure we save errno here before
calling ViceLog().

Issue spotted by kaduk@mit.edu.

Reviewed-on: https://gerrit.openafs.org/13263
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 9ff5f8f7601cc9761cc6a4ef0e8b7c8c2c8dddb5)

Change-Id: I4f41ca758574e0d58659788467372af71a5f75f2
Reviewed-on: https://gerrit.openafs.org/13898
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agovol: remove empty directories left by vos zap -force 97/13897/4
Marcio Barbosa [Sat, 11 Aug 2018 17:17:28 +0000]
vol: remove empty directories left by vos zap -force

The vos zap -force command does not remove the directories associated
with the volume in question (AFS_NAMEI_ENV). When the vos zap -force
command is executed, the volume server goes through the /vicep*/AFSIDat
directories and removes the files associated with the volume id received
as an argument. Unfortunately, the volume server does not remove the
directories associated with this volume. As a result, empty directories
are left behind.

To fix this problem, remove the empty directories left behind when vos
zap -force is executed.

Reviewed-on: https://gerrit.openafs.org/12879
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 892045a9803ed471986569705d9d727165ca7ecf)

Change-Id: I18b727a561785443f488d60b967182e3ddb9064e
Reviewed-on: https://gerrit.openafs.org/13897
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agovol: check snprintf return values in namei_ops 32/13732/5
Benjamin Kaduk [Sat, 2 Feb 2019 18:49:07 +0000]
vol: check snprintf return values in namei_ops

gcc8 is more aggressive about parsing format strings and computing bounds
on the generated text from functions like snprintf.  In this case it seems best
to detect cases of truncation and error out, rather than trying to increase
stack buffer sizes or switch to asprintf.  These paths should be well-behaved
since they are local to the fileserver, so this is mostly about appeasing the
compiler's -Wformat-truncation checks to allow us to build with --enable-checking.

Reviewed-on: https://gerrit.openafs.org/13463
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 8632f23d6718a3cd621791e82d1cf6ead8690978)

Change-Id: Ie8f9005ad9cf7cdfd3eb472e01a6fdbde5b7e57e
Reviewed-on: https://gerrit.openafs.org/13732
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agovenus: fix memory leak 95/13895/3
Marcio Barbosa [Tue, 31 May 2016 12:08:08 +0000]
venus: fix memory leak

In GetPrefCmd, when we request server prefs from the kernel and our
output buffer is not big enough, pioctl() will return E2BIG and we
allocate more memory and try again. However, if the size of the output
buffer reaches 16k bytes and this space is still not enough (or if
pioctl fails and errno != E2BIG), we return without releasing the
memory that was previously allocated.

To fix this problem, free our output buffer when this happens.

Reviewed-on: https://gerrit.openafs.org/12293
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 8ad4e15ffc883c9a99f9636d7d8a5ed0a2fcc26a)

Change-Id: I62ceddc5284c94da205ec2351ab9ef970cd64c4a
Reviewed-on: https://gerrit.openafs.org/13895
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agovenus: appease gcc8's -Wformat-string 40/13740/5
Benjamin Kaduk [Sat, 2 Feb 2019 23:09:36 +0000]
venus: appease gcc8's -Wformat-string

Interestingly, even before this commit, the buffer size was larger
than what the kernel would accept.  Since the kernel does its own
length checking, it's simplest to just allow slightly larger requests
here and have them fail later.

Reviewed-on: https://gerrit.openafs.org/13471
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit dff81f1b78fecc54f5af91f7d728925ffca62d2c)

Change-Id: Ie19d887abebdd3603a04c06723f5cb750eb654f8
Reviewed-on: https://gerrit.openafs.org/13740
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agovol: avoid query for parent id when deleting disk header 96/13896/3
Michael Meffie [Thu, 21 Dec 2017 16:59:38 +0000]
vol: avoid query for parent id when deleting disk header

When a DAFS volume server removes a volume disk header file (V*.vol),
the volume server invokes an fssync command to have the file server
delete the Volume Group Cache (VGC) entry corresponding to the volume id
and the parent id of the removed volume header.

The volume parent id is unknown to the volume server when removing a
volume disk header on behalf of a "vos zap -force" operation. In this
case, the volume server issues a fssync query to attempt look up to the
parent id from the file server's VGC.  If this fssync query fails for
some reason, volume server is unable to delete the VGC entry for the
deleted volume header. The volume server logs an error and vos zap
reports a undocumented error code.

One common way this can be encountered is to issue a "vos zap -force" on
a file server that has just been restarted. In this case, the VGC may
not be fully populated yet, so the volume server is not able to look up
the parent id of the given volume.

With this commit, relax the requirement for the parent id when deleting
VGC entries. A placeholder of 0 is used to mean any parent id for the
given volume id.

This obviates the need to query for the parent id when performing a "vos
zap -force", and allows the volume server to remove any VGC entries
associated with the volume id being zapped.

Reviewed-on: https://gerrit.openafs.org/12839
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 65b55bcc26f69f25c67518f672b34be73f3be370)

Change-Id: I2e927d7b388c7be36a67e196a3acb70e58c9a661
Reviewed-on: https://gerrit.openafs.org/13896
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Yadavendra Yadav <yadayada@in.ibm.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agovol: fix vutil format-truncation nit 31/13731/5
Benjamin Kaduk [Sat, 2 Feb 2019 18:25:35 +0000]
vol: fix vutil format-truncation nit

We need one more byte for the trailing NUL.

Reviewed-on: https://gerrit.openafs.org/13462
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit bfe912ede6f452d10cfbd5fd549f44ee027acb1b)

Change-Id: I1843bd9cb3392f721068c079b69fca65f6d1a181
Reviewed-on: https://gerrit.openafs.org/13731
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agosys: retry lsetpag if errno is EINTR 75/13975/3
Marcio Barbosa [Mon, 6 Jun 2016 17:03:54 +0000]
sys: retry lsetpag if errno is EINTR

The variable errno might be set by some system calls to indicate the
reason why the system call in question did not work as expected. If the
setpag system call is interrupted by a signal, the value of errno will
be EINTR. This value means that setpag did not succeed because it was
interrupted.

If lsetpag did not succeed and errno is equal to EINTR, try again.

Reviewed-on: https://gerrit.openafs.org/12295
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 2ae2a15c9dc9b26eaa15964cc96fdeeb6d82c74c)

Change-Id: I58d4aa633e5cadea2bc7b222f68306f07657b754
Reviewed-on: https://gerrit.openafs.org/13975
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoafs: afs_pag_wait() makes process unkillable 74/13974/3
Marcio Barbosa [Thu, 7 Nov 2019 03:10:12 +0000]
afs: afs_pag_wait() makes process unkillable

To enforce a maximum average rate of one PAG allocation per second,
afs_pag_wait(), called by afs_setpag*(), sleeps until the difference
between the current time and pag_epoch gets greater than pagCounter.
Unfortunately, this function ignores the code returned by afs_osi_Wait().
As a result, it is not possible to kill the process that requested the
new pag while afs_pag_wait() is sleeping.

To fix this problem, do not ignore the code returned by afs_osi_Wait().

Reviewed-on: https://gerrit.openafs.org/12260
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 9563807791e2402f7a214a90e96cf6ed8ea5abfb)

Change-Id: Id2453d6eb2b6cc973082da28bb3746c9f9c5ddb2
Reviewed-on: https://gerrit.openafs.org/13974
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoafs: avoid extra VL_GetEntryByName for .readonly's 68/13968/3
Marcio Barbosa [Thu, 20 Sep 2018 12:44:59 +0000]
afs: avoid extra VL_GetEntryByName for .readonly's

In the VLDB, there's only one logical entry for a volume and its
associated clones; there are not separate entries for the RW volume
"avol", the RO volume "avol.readonly", and the BK volume
"avol.backup".  And so, when looking up a volume in the VLDB by name,
the vlserver ignores any trailing ".readonly" or ".backup" in the
given name.  More concretely, the result of calling
VL_GetEntryByName*("avol") is identical to that from calling
VL_GetEntryByName*("avol.readonly").

Accordingly, if afs_GetVolumeByName(name) failed because the volume
was not found in the VLDB, afs_GetVolumeByName(name.readonly) will
fail as well (barring a change in external circumstances, such as the
volume being created or a network connection coming back up).
Therefore, the extra call in EvalMountData() is not necessary and can
be removed.

Remove the extra call, to slightly improve the response time of the
client if the volume in question does not exist, and to reduce
vlserver load when patched clients are looking up nonexistent volumes.

Reviewed-on: https://gerrit.openafs.org/13334
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 747afb94aa214217a749471679082c6ed8e81e92)

Change-Id: Ieb0bccc359fc8ebc0ad2747dbfb329d232e5e436
Reviewed-on: https://gerrit.openafs.org/13968
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agoauth: plug auth realms memory leaks 00/13900/5
Michael Meffie [Fri, 16 Nov 2018 15:00:17 +0000]
auth: plug auth realms memory leaks

The function _afsconf_FreeRealms, called by afsconf_CloseInternal, leaks
two afsconf_realms structures.

The function _afsconf_LoadRealms also leaks those two structures when it
fails.

These memory leaks were discovered with valgrind.

Reviewed-on: https://gerrit.openafs.org/13395
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
(cherry picked from commit 80ed9d98779135d43f23c9e51e7bd6bce36405f1)

Change-Id: I3e4824e2be4a22b62c1e9502860b952db777cae7
Reviewed-on: https://gerrit.openafs.org/13900
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Yadavendra Yadav <yadayada@in.ibm.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agocmd: improve help for programs without subcommands 94/13894/4
Michael Meffie [Fri, 4 Apr 2014 14:27:10 +0000]
cmd: improve help for programs without subcommands

Some programs do not have subcommands (other than the standard "help",
and "version" subcommands).  The cmd library provides the "noopcode"
mechanism for new subcommand-less programs, but older programs take
advantage of the optional "initcmd" token to simulate subcommand-less
programs.  The "initcmd" token is optional to run the command, however
it is required to display the command help.

For example, running the xstat_cm_test program without any options gives
a syntax error:

    $ xstat_cm_test
    xstat_cm_test: Missing required parameter '-cmname'
    ...

Retrying with -help (or help, -h, --help), gives the rather unhelpful
output:

    $ xstat_cm_test -help
    xstat_cm_test: Commands are:
    apropos         search by help text
    help            get help on commands
    initcmd         initialize the program

It is not obvious to the user how to get the command usage for the
program, nor that the initcmd subcommand to "initialize the program" is
actually is a placeholder to run the program.

Instead, display the command usage when help is requested and initcmd is
the only defined subcommand for a program.

For example:

    $ xstat_cm_test -help
    Usage: src/xstat/xstat_cm_test [initcmd]
             -cmname <Cache Manager name(s) to monitor>+
             -collID <Collection(s) to fetch>+ [-onceonly]
             [-frequency <poll frequency, in seconds>]
             [-period <data collection time, in minutes>] [-debug] [-help]
    Where: -onceonly  Collect results exactly once, then quit
           -debug     turn on debugging output

The libcmd library now supports an "noopcode", which should used for
future subcommand-less programs, but converting old programs to remove
the initcmd opcode could break scripts which actually specify the
optional initcmd token.

This commit adds a new libcmd flag called CMD_IMPLICIT which is used to
denote built-in subcommands such as "version" and "help".

Reviewed-on: https://gerrit.openafs.org/10983
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 77ae3dc899e89f327328c874628f100a765846c4)

Change-Id: I5b31f12f844f14e6cf31ee28c1eb60c98fcf4b59
Reviewed-on: https://gerrit.openafs.org/13894
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agorestorevol: replace snprintf with asprintf 50/13750/6
Cheyenne Wills [Fri, 2 Aug 2019 16:31:13 +0000]
restorevol: replace snprintf with asprintf

GCC is generating format-truncations warnings. With newer levels of gcc
(e.g. gcc8) and --checking-enabled these warnings result in errors and
failed builds.  In addition clang8 static analysis tools are reporting
memory leaks.

Replace snprintf with asprintf and eliminate some of the large work
buffers that are being placed on the stack. In order to correct some of
the format-truncation errors the size of the buffers grew significantly
(e.g. gcc is reporting the need to resize some of the buffers from 256
bytes to 4K in order to eliminate the warnings).

Ensure allocated work buffers are freed before function return.

Obtained a clean build with gcc9/clang8 with --enable-checking and a
clean scan-build report with clang8.

Reviewed-on: https://gerrit.openafs.org/13494
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit bf24b301a10dcb5710a98e58252213bd72c6f352)

Change-Id: If9fa37613841ffd090ec565dc24171bf89579c5b
Reviewed-on: https://gerrit.openafs.org/13750
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

3 months agobos: remove smail-notifier 38/13738/7
Cheyenne Wills [Fri, 1 Mar 2019 15:46:32 +0000]
bos: remove smail-notifier

smail-notifier is a sample program that is undocumented and has not
been well maintained.  It produces copious compiler warnings, and
would require effort to bring the code up to decent coding practices.

The bosserver provides a -notifier feature that can be used for
notifications, but that feature does not depend on this sample program.

Removed the code, cleaned up the Makefiles and .gitignore.

Reviewed-on: https://gerrit.openafs.org/13509
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 6e988a5b3900fe73c314c9960d6fb7753ff98411)

Change-Id: I073a2b772f894e321bd0b41e012229c8e6d3105c
Reviewed-on: https://gerrit.openafs.org/13738
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoLINUX 5.3: Add comments for fallthrough switch cases 10/13910/4
Cheyenne Wills [Tue, 1 Oct 2019 18:14:41 +0000]
LINUX 5.3: Add comments for fallthrough switch cases

With commit 6e0f1c3b45102e7644d25cf34395ca980414317f (LINUX: Honor
--enable-checking for libafs) building libafs against a linux 5.3
kernel compiles with errors due to fall through in case statements when
--enable-checking / --enable-warning is used.

e.g.
  src/opr/jhash.h:82:17: error: this statement may fall through
                                [-Werror=implicit-fallthrough=]
         case 3 : c+=k[2];
                  ~^~~~~~

The GCC compiler will disable the implicit-fallthrough check for case
statements that contain a "special" comment ( /* fall through */ ).

Add the 'fall through' comment to indicate where fall throughs are
acceptable.

This commit only adds comments and does not alter any executable code.

The -Wimplicit-fallthrough flag was enabled globally in the linux kernel
build in 5.3-rc2 (commit: a035d552a93bb9ef6048733bb9f2a0dc857ff869
Makefile: Globally enable fall-through warning)

Reviewed-on: https://gerrit.openafs.org/13881
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit a455452d7ee98d160620925bb8a0e3d0f4dfd7ec)

Change-Id: Icad4b5923d971e7519f5d5259cd9c009c40c0d7a
Reviewed-on: https://gerrit.openafs.org/13910
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoLINUX: Unlock page on afs_linux_read_cache errors 65/13765/3
Andrew Deason [Wed, 3 Jul 2019 17:55:53 +0000]
LINUX: Unlock page on afs_linux_read_cache errors

When afs_linux_read_cache is called with a non-NULL task, it is
responsible for unlocking 'page' (unless it's unlocked in a background
task), even if we encounter an error. Currently we almost always do
unlock the given page for a non-NULL task, but if we manage to hit one
of the codepaths that 'goto out', we skip over the unlock_page() call
near the end of the function, and the page never gets unlocked.

As a result, the page stays locked forever. That generally means any
future access to the same file will block forever, and when we try to
flush the relevant vcache, we will block waiting for the page lock
while holding GLOCK. (This can happen via the background daemon via
e.g. afs_ShakeLooseVCaches -> osi_TryEvictVCache -> afs_FlushVCache ->
osi_VM_FlushVCache -> vmtruncate -> ... -> truncate_inode_pages_range
-> __lock_page on Linux 2.6.32-754.2.1.el6.) This quickly brings the
whole client to a halt until the machine can be forcibly rebooted.

To solve this, just move the 'out:' label to before the page unlock.
Add a few locking-related comments around the relevant code to help
explain some relevant details.

The relevant code has changed and been refactored over the years, but
this problem has probably existed ever since this code was originally
converted to using the readpage() of the underlying cache fs, in
commit 88a03758 (Use readpage, not read for fastpath access).

Reviewed-on: https://gerrit.openafs.org/13672
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit eed79e2d28dcab889d01869e57dec14fd30d421c)

Change-Id: I6391897473e701bd81eb334935317dc5009612da
Reviewed-on: https://gerrit.openafs.org/13765
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoLINUX: Avoid re-taking global lock in afs_dentry_iput 48/13748/3
Yadavendra Yadav [Fri, 26 Jul 2019 14:29:25 +0000]
LINUX: Avoid re-taking global lock in afs_dentry_iput

“dput” function internally can call dentry_iput which  results in
calling afs_dentry_iput. So in case before calling “dput” if global lock
was held then when afs_dentry_iput is called it will again try to lock
global lock and will result in deadlock scenario.  So to avoid this
deadlock make sure if global lock is already taken before calling
afs_dentry_iput, don’t try to lock it again.  This issue was partially
fixed in commit 0dac4de8 (Linux: drop GLOCK before calling dput)

Reviewed-on: https://gerrit.openafs.org/13725
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 5792e0211be275cf79d10e8c5f6ab2a14493e07a)

Change-Id: I4a17700adb18956fc61462663fdb690b267cc928
Reviewed-on: https://gerrit.openafs.org/13748
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoafs: Free 'addrs' array 99/13899/5
Andrew Deason [Mon, 1 Oct 2018 15:56:53 +0000]
afs: Free 'addrs' array

Currently, 3 places in libafs allocate an 'addrs' array in a very
similar way to loop through our list of servers:
ForceAllNewConnections(), afs_LoopServers(), and PCallBackAddr(). Of
these, only afs_LoopServers actually frees the array.
ForceAllNewConnections and PCallBackAddr leak the memory, but these
are only hit from infrequent pioctls that can only be run by root, so
the impact is small.

Fix ForceAllNewConnections and PCallBackAddr to free the array.

Reviewed-on: https://gerrit.openafs.org/13355
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 0548ee436d0f0f92a980d22e03149faedf38dc70)

Change-Id: I5d64899c7be40ba3e1b0985c4829933eebbd8323
Reviewed-on: https://gerrit.openafs.org/13899
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agorx: Fix test for end of call queue for LWP 92/13892/3
Andrew Deason [Thu, 26 Sep 2019 18:35:51 +0000]
rx: Fix test for end of call queue for LWP

Commit 6ad3d646 (rx: Correctly test for end of call queue) fixed a
broken end-of-queue check in rx_GetCall, but it only fixed the
RX_ENABLE_LOCKS version of rx_GetCall. The non-locks version (i.e. the
LWP version) still had this bug. Fix it for the LWP case, to avoid
some rare cases where an Rx call can get stuck in the incoming queue.

Also remove the comment added by commit 170dbb3c (rx: Use opr queues),
since we're fixing the mentioned problem.

Reviewed-on: https://gerrit.openafs.org/13880
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit d9fc4890f01a41fa5a63f97f2446b3afc35b473f)

Change-Id: I2e0106b63a8bf09634500944490dfae2e86c18b9
Reviewed-on: https://gerrit.openafs.org/13892
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoThe interminable rework of afs_random() 79/13879/4
Benjamin Kaduk [Fri, 9 Aug 2019 14:59:44 +0000]
The interminable rework of afs_random()

Commit f0a3d477d6109697645cfdcc17617b502349d91b restructured the
operation on tv_usec to avoid using undefined behavior, but in
the process introduced a behavior change.  Historically (at least as
far back as AFS-3.3), we masked off the low nybble (four bits) of
tv_usec before adding the low byte (eight bits) of the rxi_getaddr()
output.  Why there was a desire to combine two sources of input for
the overlapping four bits remains unclear, but restore the historical
behavior for now, as the intent of commit
f0a3d477d6109697645cfdcc17617b502349d91b was to not introduce any
behavior changes.

Reviewed-on: https://gerrit.openafs.org/13759
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 1c4e94da2a8fce9d79006ad6d6673d3d7de117d3)

Change-Id: Iec10673e5ec73c1e0edcc231690cb6133fce8691
Reviewed-on: https://gerrit.openafs.org/13879
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agorework afs_random() yet again 43/13743/5
Benjamin Kaduk [Sun, 3 Feb 2019 01:45:31 +0000]
rework afs_random() yet again

clang 7 notes that ~0 is signed and that left-shifting into the sign
bit is undefined behvaior.  Use a new construction to clear the low
byte of tv_usec with only bitwise operations that are independent of
the width of tv_usec and stay within the realm of C's defined behavior.

Reviewed-on: https://gerrit.openafs.org/13474
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit f0a3d477d6109697645cfdcc17617b502349d91b)

Change-Id: I4f0438c2fc8237968f41409ca23ac098839508fe
Reviewed-on: https://gerrit.openafs.org/13743
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoAvoid incomplete function type in casts 42/13742/5
Benjamin Kaduk [Sun, 3 Feb 2019 00:39:53 +0000]
Avoid incomplete function type in casts

clang complains that these casts contain an incomplete function type
(since the function argument is omitted rather than declared to be
void).  Since we just need the cast to pointer type, let the compiler
do it implicitly and pass stock NULL, rather than trying to force a
cast to function-pointer type.

Reviewed-on: https://gerrit.openafs.org/13473
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 96c0b88947c7aab605170bdca633d3716051a58e)

Change-Id: I950ff8de925a1ca03e50ad7ec394123445b5ce4a
Reviewed-on: https://gerrit.openafs.org/13742
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoAvoid calling krb5_free_context(NULL) 02/13902/5
Andrew Deason [Fri, 1 Feb 2019 22:31:50 +0000]
Avoid calling krb5_free_context(NULL)

Several places in the code currently call krb5_free_context(ctx) in a
cleanup code path, where 'ctx' may or may not be NULL. This is not
guaranteed to be okay, so check for NULL to make sure we don't cause
issues in these code paths.

While we are here cleaning up krb5_free_context() calls, also fix a
few call sites in afscp_util.c that were not calling krb5_free_context
in all error paths.

Reviewed-on: https://gerrit.openafs.org/13461
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 26b1dc036719a588a5cadecb14053bd4079c1f48)

Change-Id: I3b0d22f51f4fe85897116b7f96d096570258eed2
Reviewed-on: https://gerrit.openafs.org/13902
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agovlserver: initialize nvlentry elements after read 46/13846/3
Cheyenne Wills [Fri, 9 Aug 2019 19:25:26 +0000]
vlserver: initialize nvlentry elements after read

Commit 7620bd33487207b348ed7aeba45f8d743132ba84 (vlserver: fix
vlentryread() for old vldb formats) leaves the tail end of the
serverNumber, serverParition and serverFlags arrays uninitialized since
it only copies OMAXNSERVERS elements into arrays that have NMAXNSERVERS
elements.

Initialize the elements in the nvlentry server arrays that were not
copied with BADSERVERID.

Reviewed-on: https://gerrit.openafs.org/13755
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit ddf7d2a7f4bfdcab238e791cb8c49bb803e76b09)

Change-Id: I4e1065bedda0f50b85cf472d015f2c86e4af82c8
Reviewed-on: https://gerrit.openafs.org/13846
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agovlserver: fix vlentryread() for old vldb formats 34/13734/4
Benjamin Kaduk [Sat, 2 Feb 2019 20:23:03 +0000]
vlserver: fix vlentryread() for old vldb formats

When we're using old format compatibility, use OMAXNSERVERS for the
array lengths instead of MAXNSERVERS.  Otherwise we'll try to copy more
data than we've read.

Detected by gcc8 as:

vlutils.c:183:2: error: ‘memcpy’ forming offset [149, 151] is out of the bounds [0, 148] of object ‘tentry’ with type ‘struct vlentry’ [-Werror=array-bounds]
  memcpy(nbufp->serverFlags, oep->serverFlags, NMAXNSERVERS);
  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
vlutils.c:141:26: note: ‘tentry’ declared here
     struct vlentry *oep, tentry;
                          ^~~~~~

Reviewed-on: https://gerrit.openafs.org/13465
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 7620bd33487207b348ed7aeba45f8d743132ba84)

Change-Id: I7dc4ad48805c6a82dd021d156fe187dd97e5b456
Reviewed-on: https://gerrit.openafs.org/13734
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agouss: Allocate buffer space for trailing NUL 36/13736/4
Benjamin Kaduk [Sat, 2 Feb 2019 21:26:23 +0000]
uss: Allocate buffer space for trailing NUL

Appease gcc8's -Wformat-truncation engine.

Reviewed-on: https://gerrit.openafs.org/13467
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 9a5ba85d1853327d8184287e58a6e03fabaaf23d)

Change-Id: I9900452b951d600e6358b216015fe4fcdf8f633e
Reviewed-on: https://gerrit.openafs.org/13736
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agouss: uss_procs.c format-overflow warning 29/13729/4
Cheyenne Wills [Fri, 26 Jul 2019 13:59:33 +0000]
uss: uss_procs.c format-overflow warning

GCC 9 introduced new warnings/errors and is flagging a sprintf with a
format-overflow warning.  With --checking-enabled, this error is causing
uss_procs.c to fail during compile.

A file name with the full path is being composed and the size of the
buffer was triggering a possible format-overflow warning/error.

Use asprintf to allocate the buffer dynamically instead of using a
buffer sitting on the stack (reducing the stack requirements by 2K).

Produces new error message if asprintf returns an error.

Reviewed-on: https://gerrit.openafs.org/13664
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 41ee558329560bce037ad2860282d8b49aa11b2d)

Change-Id: I5c5866142ae17c92017201fb567f847b5c2907a0
Reviewed-on: https://gerrit.openafs.org/13729
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoptserver: testpt.c format-overflow warning 30/13730/4
Cheyenne Wills [Tue, 2 Jul 2019 22:58:28 +0000]
ptserver: testpt.c format-overflow warning

GCC 9 introduced new warnings/errors and is flagging a sprintf with a
format-overflow warning.  With --checking-enabled, this error is causing
testpt.c to fail during compile.

Change the buffer size from 16 bytes to PR_MAXNAMELEN+1 and use snprintf
instead of sprintf. Generate an error message and exit if snprintf
truncates the string.

Reviewed-on: https://gerrit.openafs.org/13663
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 4a57cc54dfb6789a86ee735360ee44209c1a901a)

Change-Id: I2f8012e7fb4384f3ad877d2c9beb5f00b03716b8
Reviewed-on: https://gerrit.openafs.org/13730
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoptserver: Incorrect variable used to print error msg 28/13728/4
Cheyenne Wills [Tue, 25 Jun 2019 21:39:40 +0000]
ptserver: Incorrect variable used to print error msg

In testpt.c the variable cdir is used to print the name of the temporary
dir.  However at this point in the code cdir is NULL and the variable
tmp_conf_dir contains the actual name that should be used in the error
message.

Flagged as an error when --enable-checking is on and using GCC 9.

Reviewed-on: https://gerrit.openafs.org/13662
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit f938f5f248a3cb3f7ac871f5ef45a0e2d043706b)

Change-Id: I1b993ddc2545f90736811e2eb85ba4b3bae6e657
Reviewed-on: https://gerrit.openafs.org/13728
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoaklog: require opt-in to enable single-DES in libkrb5 91/13791/2
Benjamin Kaduk [Fri, 12 Jul 2019 04:07:35 +0000]
aklog: require opt-in to enable single-DES in libkrb5

Since the introduction of rxkad-k5 in response to OPENAFS-SA-2013-003,
it is not strictly necessary to configure libkrb5 to allow weak crypto
in order to obtain an AFS token.  A sufficient amount of time has passed
since then that it is safe to assume that the default behavior is the
more-secure one, and require opt-in for the insecure behavior.

To indicate that the use of single-DES is quite risky, add the
"-insecure_des" argument to both klog and aklog, to gate the
preexisting calls that enable weak crypto/single-DES.
These calls, and the -insecure_des option, may be removed entirely
in a future commit.

Reviewed-on: https://gerrit.openafs.org/13689
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit eaae6eba8ca10ba7a5a20ee0d1b5f91bc2bac6c6)

Change-Id: I197042e12567fa0fed1b6584e85c3f0a520efa4c
Reviewed-on: https://gerrit.openafs.org/13791
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agorxkad: ticket5.c fix typo in #if statement 56/13756/3
Cheyenne Wills [Thu, 8 Aug 2019 18:07:51 +0000]
rxkad: ticket5.c fix typo in #if statement

commit 98ca332c4a5ac9e5687fb4fe21b350134bc74d1b (rxkad: v5der.c format
truncation warnings) contains a typo in the test for clang (_clang
instead of __clang__)

Correct the typo in the #if statement to test for __clang__

Reviewed-on: https://gerrit.openafs.org/13754
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit e3dbd8a5886734f6390126e155cc259b0de5af51)

Change-Id: I6c2a9851e668947c61d97e013c84808f738c4e52
Reviewed-on: https://gerrit.openafs.org/13756
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agorxkad: v5der.c format truncation warnings 27/13727/4
Cheyenne Wills [Mon, 15 Jul 2019 14:38:24 +0000]
rxkad: v5der.c format truncation warnings

GCC 7 is producing new warnings due to better compile time analysis.
With --enable-checking v5der.c is failing with 2 errors due to possible
format-truncation in some snprintf calls.  The format strings are being
used to format a date and time values from a tm structure.

The actual warnings/errors are being triggered from arithmetic being
performed on the year and month members of the structure. The resulting
values should not exceed the format lengths, but the compilers are still
flagging the statements.

v5der.c is part of the heimdal package that is pulled into the openafs
source tree.  v5der.c is not compiled directly but is #included in
ticket5.c

Update ticket5.c to change the severity of the format-truncation
diagnostic to a warning if using GCC 7 (or higher).

Note: since v5der.c is pulled from an external source (heimdal), any
changes to update v5der.c directly would need to be performed upstream.

Reviewed-on: https://gerrit.openafs.org/13661
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 98ca332c4a5ac9e5687fb4fe21b350134bc74d1b)

Change-Id: I1a808060b302549887e529e74bc3805d9431c499
Reviewed-on: https://gerrit.openafs.org/13727
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoptserver: Increase length limit of namelist, idlist, prlist, prentries 44/13844/4
Stephan Wiesand [Fri, 6 Sep 2019 11:35:02 +0000]
ptserver: Increase length limit of namelist, idlist, prlist, prentries

An implementation limit of those lists was introduced in commit
a0ffea098d8c5c5b46c6bf86a12d28d6e7096685 to prevent using unlimited
amounts of memory in ptserver and the client.  Subsequent reports
indicate that the chosen limits are small enough to restrict
functionality currently in use at some sites where membership lists
exceed the current limit.  Since this is just an implementation-
defined limit and can freely change from release to release, increase
the threshold by an order of magnitude to preserve functionality for
existing deployments while still retaining some protection against
attacker-controlled excessive memory allocation.

Reviewed-on: https://gerrit.openafs.org/13838
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit d1e90b82ebb2685cbac3ecb3fd99136328b35357)

Change-Id: Ifa229179ad6d2962a8d49df6abd1add94fad7259
Reviewed-on: https://gerrit.openafs.org/13844
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoWINNT: Link tbutc against mtafsutil.lib 52/13852/5
Andrew Deason [Tue, 27 Aug 2019 01:33:58 +0000]
WINNT: Link tbutc against mtafsutil.lib

tbutc uses pthreads, not LWP, so link it against mtafsutil.lib (a
pthread library), and not afsutil.lib (an LWP library).

Reviewed-on: https://gerrit.openafs.org/13822
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 7a76f4dc00984d42b0535a8edbedee034ada896f)

Change-Id: I133fff53d1974658ed1fe95e48abd9779a346a4f
Reviewed-on: https://gerrit.openafs.org/13852
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agorx: Export rx_GetCallStatus 51/13851/5
Andrew Deason [Tue, 27 Aug 2019 00:34:19 +0000]
rx: Export rx_GetCallStatus

Commit 59d3a8b8 (vos: restore status information to 'vos status')
added the function rx_GetCallStatus to Rx, and used it in the
volserver, but didn't add the function to our .sym and .exp files,
causing a linker error on at least WINNT.

Add the function to the relevant .sym/.exp files, so we can link on
all platforms.

Change-Id: I859ac6d04d8a21eb6f8b4ba3f3720ca318e91334
Reviewed-on: https://gerrit.openafs.org/13820
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit c3716b3d7e32f47b084657e163b029e9f1756fa4)
Reviewed-on: https://gerrit.openafs.org/13851
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>
Tested-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoWINNT: Link butc against audit 50/13850/5
Andrew Deason [Mon, 26 Aug 2019 23:14:48 +0000]
WINNT: Link butc against audit

Since commit c43169fd (OPENAFS-SA-2018-001 Add auditing to butc server
RPC implementations), butc references symbols from audit. So add audit
to our libraries to link against, so we can link butc on WINNT.

Reviewed-on: https://gerrit.openafs.org/13818
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit e4b689e8c7cb39b72854dd38b6a92134591c8bca)

Change-Id: Ib27755730178afbbd85e3aad265c1f956b3785ef
Reviewed-on: https://gerrit.openafs.org/13850
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>
Tested-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoafs: Avoid giving wrong 'tf' to afs_InitVolSlot 37/13937/2
Andrew Deason [Tue, 5 Nov 2019 16:50:01 +0000]
afs: Avoid giving wrong 'tf' to afs_InitVolSlot

Commit 75e3a589 (libafs: afs_InitVolSlot function) split out a bit of
our code that initializes a struct volume into the afs_InitVolSlot
function. However, it caused us to almost always pass a non-NULL 'tf'
to afs_InitVolSlot, even if the target volume was not found.

That is, before that commit, our code roughly did this:

    for (...; j != 0; j = tf->next) {
        ...;
        tf = &staticVolume;
        if (tf->volume == volid)
            break;
    }
    if (tf && j != 0) {
        use_tf_data();
    } else {
        use_blank_data();
    }

The reason for the extra 'j != 0' check after the loop is to see if we
hit the end of the volume hash chain, or if we actually found a
matching 'tf' in the loop.

And after that commit, the code did this:

    for (...; j != 0; j = tf->next) {
        ...;
        if (j != 0) {
            tf = &staticVolume;
            if (tf->volume == volid)
                break;
        }
    }
    if (tf) {
        use_tf_data();
    } else {
        use_blank_data();
    }

The check for 'j != 0' was moved to inside the for loop, but 'j' is
always nonzero in the loop (otherwise, the for() would exit the loop).
This means that if we didn't find a matching 'tf' in the loop, our
'tf' would be non-NULL anyway, and so we'd initialize our volume slot
from just the last entry in the hash chain.

This means that for volumes that are not found in the VolumeItems
file, our struct volume will probably be initialized with arbitrary
data from another volume, instead of being initialized to the normal
defaults (the 'else' clause in afs_InitVolSlot).

This means that the 'dotdot' entry for the volume may be wrong, and so
we may report the wrong parent dir for the root of a volume. However,
the 'dotdot' entry should be fixed when the volume root is accessed
via a mountpoint, so any such issue should be temporary. And of
course, on some platforms (LINUX) we don't ever use the 'dotdot'
information for a volume, and even on other platforms, often resolving
the '..' entry is handled by other means (e.g. shells often calculate
it themselves). But some 'pwd' calculations and other '..' corner
cases may be affected.

To fix this, change the relevant loop so that we only set 'tf' to
non-NULL when we actually find a matching entry.

Reviewed-on: https://gerrit.openafs.org/13933
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 4a9078c6bbf51720a5eacf7e6ba21443e5103eee)

Change-Id: Ib1e7519db8f844872c4b88b54978f358ff7b299e
Reviewed-on: https://gerrit.openafs.org/13937
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoviced: consistently enforce host thread quota for ICBS(3) 93/13893/2
Mark Vitale [Tue, 17 Sep 2019 19:14:44 +0000]
viced: consistently enforce host thread quota for ICBS(3)

From time to time, the fileserver may issue potentially long-running
RXAFSCB_* RPCs back to a host (client).  If these are holding h_Lock_r
(host->lock) while running, they may cause other service threads for the
same host (client) to block.

In order to prevent a given host from tying up too many service threads
in this way, the fileserver enforces a quota limiting how many threads
can be waiting for h_Lock_r on a particular host while waiting for one
of the following RPCs to complete:
- RXAFSCB_TellMeABoutYourself (TMAY)
- RXAFSCB_WhoAreYou
- RXAFSCB_ProbeUuid
- RXAFSCB_InitCallBackState (ICBS)
- RXAFSCB_InitCallBackState3 (ICBS3)

Note: Although some of these RPCs are relatively lightweight, they may
still experience network delays.

This quota is enforced by calling h_threadquota() in h_Lookup_r and
h_GetHost_r.  The quota check is enabled for a given host by turning on
host->hostFlags HWHO_INPROGRESS for the duration of the RXAFSCB_* RPC.
The quota check is only needed, and should only be enabled, when the RPC
is issued while h_Lock_r is held.

However, there are a few paths to ICBS(3) where h_Lock_r is held but
HWHO_INPROGRESS is not set.  A delay in those paths may allow a host to
consume an unlimited number of fileserver threads.  One such path
observed in a field report was SRXAFS_FetchStatus -> CallPreamble ->
BreakDelayedCallBacks_r -> RXAFSCB_ICBS3.

Instead, enable host thread quotas for all remaining unregulated ICBS(3)
RPCs.

Reviewed-on: https://gerrit.openafs.org/13873
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit aefc4c4f46e13f59b4cbe043e1a2a6f4ed99e076)

Change-Id: If3883a152078bba9995e0c8a13ab31788db6347f
Reviewed-on: https://gerrit.openafs.org/13893
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>
Tested-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agokauth: Move COUNT_REQ to beginning of block 49/13849/5
Andrew Deason [Mon, 26 Aug 2019 21:08:31 +0000]
kauth: Move COUNT_REQ to beginning of block

Commit b604ee7a (OPENAFS-SA-2018-002 kaserver: prevent KAM_ListEntry
information leak) added a memset in kamListEntry before COUNT_REQ, but
COUNT_REQ declares a local variable. This breaks the WINNT build,
because we must declare variables at the beginning of a block.

To fix this, just swap the two lines.

Reviewed-on: https://gerrit.openafs.org/13815
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit b9b5385e6a04dcacd180f33e39495c7909fe4df3)

Change-Id: Id9c1fd67e4614f8f433415486e107ecb4bd0d708
Reviewed-on: https://gerrit.openafs.org/13849
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>
Tested-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoWINNT: Build bubasics before audit 48/13848/5
Andrew Deason [Mon, 26 Aug 2019 18:13:28 +0000]
WINNT: Build bubasics before audit

Commit 9ebff4c6 (OPENAFS-SA-2018-001 audit: support butc types) made
src/audit require the butc.h header, and updated Makefile.in to
reflect this. However, this dir is also built on WINNT, and the
NTMakefile was not updated to reflect this dependency. As a result, we
might fail to build src/audit on WINNT, since butc.h may not exist
yet, and we get an error like:

            cl [...] /c audit.c
    audit.c
    cl : Command line warning D9025 : overriding '/W4' with '/W3'
    audit.c(27) : fatal error C1083: Cannot open include file: 'afs/butc.h': No such file or directory
    NMAKE : fatal error U1077: 'C:\PROGRA~2\MICROS~1.0\VC\bin\amd64\cl.EXE' : return code '0x2'

To fix this, move 'bubasics' to be made before 'audit' in NTMakefile,
so butc.h is available when we build 'audit'.

Reviewed-on: https://gerrit.openafs.org/13813
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 9eeb3ec09f5421ceab2be415a193bb3a3c44925f)

Change-Id: If36de5664ea0eb7208810c224d30092f0a4d1745
Reviewed-on: https://gerrit.openafs.org/13848
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>
Tested-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoafs: Avoid panics in afs_InvalidateAllSegments 47/13847/2
Andrew Deason [Mon, 8 Jul 2019 19:49:23 +0000]
afs: Avoid panics in afs_InvalidateAllSegments

Currently, afs_InvalidateAllSegments panics when afs_GetValidDSlot
fails. We panic in these cases because afs_InvalidateAllSegments
cannot simply return an error to its callers; we must invalidate all
segments for the given vcache, or we risk serving incorrect data to
userspace as explained in the comments.

Instead of panicing, though, we could simply sleep and retry the
operation until it succeeds. Implement this, retrying every 10
seconds, and logging a message every hour that we're stuck (in case
we're stuck for a long time).

When we retry the operation, do so in a background request, to avoid a
somewhat common situation on Linux where we always get I/O errors from
the cache when the calling process has a SIGKILL pending. Create a new
background op for this, BOP_INVALIDATE_SEGMENTS.

With this, the relevant vcache will be effectively unusable for the
entire time we're stuck in this situation (avc->lock will be
write-locked), but this is at least better than panicing the whole
machine.

Reviewed-on: https://gerrit.openafs.org/13677
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 3be5880d1d2a0aef6600047ed43d602949cd5f4d)

Change-Id: Iba1cde70a4d5e919fedfe27d0540878113a369e4
Reviewed-on: https://gerrit.openafs.org/13847
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agogtx: Avoid incomplete function type in casts 49/13749/2
Cheyenne Wills [Fri, 26 Jul 2019 20:57:02 +0000]
gtx: Avoid incomplete function type in casts

clang complains that these casts contain an incomplete function type
(since the function argument is omitted rather than declared to be
void).  Since we just need the cast to pointer type, let the compiler
do it implicitly and pass stock NULL, rather than trying to force a
cast to function-pointer type.

Reviewed-on: https://gerrit.openafs.org/13726
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit d6262c3f391e4176bec207fd0e8d4d6091a7f4e2)

Change-Id: I4544c37591bb68ff6bbe345192490bb79c843fc5
Reviewed-on: https://gerrit.openafs.org/13749
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agolibadmin: appease clang -Wsometimes-uninitialized 45/13745/4
Benjamin Kaduk [Sun, 3 Feb 2019 01:52:26 +0000]
libadmin: appease clang -Wsometimes-uninitialized

clang thinks that 'time' can be used uninitialized:

bos.c:1472:9: error: variable 'time' is used uninitialized whenever 'if' condition is
      false [-Werror,-Wsometimes-uninitialized]
    if (as->parms[TIME].items) {
        ^~~~~~~~~~~~~~~~~~~~~
bos.c:1478:57: note: uninitialized use occurs here
    if (!bos_ExecutableRestartTimeSet(bos_server, type, time, &st)) {
                                                        ^~~~
bos.c:1472:5: note: remove the 'if' if its condition is always true
    if (as->parms[TIME].items) {
    ^~~~~~~~~~~~~~~~~~~~~~~~~~~
bos.c:1445:5: note: variable 'time' is declared here
    bos_RestartTime_t time;
    ^

but in this command description, the TIME argument is required.
Add a never-triggered error exit to appease the compiler when
--enable-checking is activated.

Reviewed-on: https://gerrit.openafs.org/13476
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 453060c27a5d33d3c27128d169298f9d66d06f1a)

Change-Id: Iac80d4ec7c2a33dcb470de2daedf693c20b96b00
Reviewed-on: https://gerrit.openafs.org/13745
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agouss: signed/unsigned char fallout 44/13744/4
Benjamin Kaduk [Sun, 3 Feb 2019 01:48:20 +0000]
uss: signed/unsigned char fallout

When char is signed, assigning 255 to a variable of type char changes
the value, which causes clang to emit a warning and fail the
--enable-checking build.

Reviewed-on: https://gerrit.openafs.org/13475
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 7c15e6efe62fb3fe1970c56331df09b257abf6d9)

Change-Id: I3dd374582b57e46460ea80ead75913948c2d2262
Reviewed-on: https://gerrit.openafs.org/13744
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agodumpscan: appease gcc8 -Wformat-overflow 41/13741/4
Benjamin Kaduk [Sat, 2 Feb 2019 23:10:29 +0000]
dumpscan: appease gcc8 -Wformat-overflow

gcc does not benefit from our external knowledge that tm_year is
tightly bounded, and thinks it could still be in the range
[-21474817482147483647], which would overflow our string buffer.
The function in question does not have error handling in place, so
rather than adding some or trying to assert the proper bounds, just
use a slightly larger buffer for safety.

Reviewed-on: https://gerrit.openafs.org/13472
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 8f03ff3bdd8eb9f4557cdb7054aee9b8ea432160)

Change-Id: I05c8d998c6d40118a1bde923e346cddbdfa4192b
Reviewed-on: https://gerrit.openafs.org/13741
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agoscout: band-aid -Wformat-truncation 39/13739/4
Benjamin Kaduk [Sat, 2 Feb 2019 23:02:08 +0000]
scout: band-aid -Wformat-truncation

gcc8 gets pretty confused about the bounds on these things (presumably
due to our alignment options) and thinks this could potentially be a huge
string.  Check for truncation to appease the compiler, instead of trying
to ensure that the buffer is big enough.

Reviewed-on: https://gerrit.openafs.org/13470
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit df8534909fdc1fa8417aa788c0fa71c5dbe7eb30)

Change-Id: Idf3a2f32ba4630a7d11b2c0664c6dd9b694eb7db
Reviewed-on: https://gerrit.openafs.org/13739
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agobutc: -Wformat-truncation fallout 37/13737/4
Benjamin Kaduk [Sat, 2 Feb 2019 21:44:54 +0000]
butc: -Wformat-truncation fallout

Increase some buffer sizes to appease gcc8.  While here, use snprintf
instead of plain sprintf(!).

Reviewed-on: https://gerrit.openafs.org/13468
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit a89297a066d8689f8fc29a7428cfe3ed6235d010)

Change-Id: If26312359d62649077e1f8adf2c9e0b39d28496f
Reviewed-on: https://gerrit.openafs.org/13737
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agovlserver: use large enough buffer for rxinfo string 35/13735/4
Benjamin Kaduk [Sat, 2 Feb 2019 20:43:04 +0000]
vlserver: use large enough buffer for rxinfo string

The "[dotted-quad] rxkad:name.inst@cell" construct can be as large as
(3*4+3)+7+3*64+2+1 == 217 characters (including trailing NUL); size
our buffer accordingly to avoid the risk of truncation.

Reviewed-on: https://gerrit.openafs.org/13466
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 584b0f2b6b4391c0c879352bb1786c0f267666c9)

Change-Id: Ia11e685ec17f34a9a8fdc42d392b8a2677f63696
Reviewed-on: https://gerrit.openafs.org/13735
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

4 months agovol: avoid -Wformat-truncation issues in vol-salvage.c 33/13733/4
Benjamin Kaduk [Sat, 2 Feb 2019 18:56:26 +0000]
vol: avoid -Wformat-truncation issues in vol-salvage.c

Make some formerly-64-character buffers VMAXPATHLEN (plus a smidgeon)
to give them space to hold the composed paths.

Reviewed-on: https://gerrit.openafs.org/13464
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit d6b88e3bd5219a8dffebc07df23e30f1d16f095f)

Change-Id: Idd388cd23ffeff1307b9eb6d31976383b6125260
Reviewed-on: https://gerrit.openafs.org/13733
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Marcio Brito Barbosa <mbarbosa@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

7 months agoMake OpenAFS 1.8.5 20/13920/2 openafs-stable-1_8_5
Benjamin Kaduk [Tue, 22 Oct 2019 07:11:09 +0000]
Make OpenAFS 1.8.5

Update version strings for the 1.8.5 release.

Change-Id: I6e4e4b02b2ad7686027e983d63919cd3045fd2d4
Reviewed-on: https://gerrit.openafs.org/13920
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>

7 months agoUpdate NEWS for 1.8.5 19/13919/3
Benjamin Kaduk [Tue, 22 Oct 2019 07:08:36 +0000]
Update NEWS for 1.8.5

Release notes for the OpenAFS 1.8.5 security release.

Change-Id: Idd44efa17c41a9fa4d2d3beddb294a1c24bdec9e
Reviewed-on: https://gerrit.openafs.org/13919
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>

7 months agoOPENAFS-SA-2019-003: ubik: Avoid unlocked ubik_currentTrans deref 18/13918/2
Andrew Deason [Mon, 16 Sep 2019 19:06:53 +0000]
OPENAFS-SA-2019-003: ubik: Avoid unlocked ubik_currentTrans deref

Currently, SVOTE_Debug/SVOTE_DebugOld examine some ubik internal state
without any locks, because the speed of these functions is more
important than accuracy.

However, one of the pieces of data we examine is ubik_currentTrans,
which we dereference to get ubik_currentTrans->type. ubik_currentTrans
could be set to NULL while this code is running, so there is a small
chance of this code causing a segfault, if SVOTE_Debug() is running
when the current transaction ends.

We only ever initialize ubik_currentTrans as a write transation (via
SDISK_Begin), so this check is pointless anyway. Accordingly, skip the
type check, and always assume that any active transaction is a write
transaction.  This means we only ever access ubik_currentTrans once,
avoiding any risk of the value changing between accesses (and we no
longer need to dereference it, anyway).

Note that, since ubik_currentTrans is not marked as 'volatile', some C
compilers, with certain options, can and do assume that its value will
not change between accesses, and thus only fetch the pointer value once.
This avoids the risk of NULL dereference (and thus, crash, if pointer
stores/loads are atomic), but the value pointed to by ubik_currentTrans->type
would be incorrect when the transaction ends during the execution of
SVOTE_Debug().

Reviewed-on: https://gerrit.openafs.org/13915
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 6ec46ba7773089e1549d27a0d345afeca65c9472)

Change-Id: I634ddb27e7a8dbe5c9d1dacdc83070efa470b50b
Reviewed-on: https://gerrit.openafs.org/13918
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>

7 months agoOPENAFS-SA-2019-002: Zero all server RPC args 17/13917/3
Andrew Deason [Thu, 8 Aug 2019 02:19:47 +0000]
OPENAFS-SA-2019-002: Zero all server RPC args

Currently, our server-side RPC argument-handling code generated from
rxgen initializes complex arguments like so (for example, in
_RXAFS_BulkStatus):

    AFSCBFids FidsArray;
    AFSBulkStats StatArray;
    AFSCBs CBArray;
    AFSVolSync Sync;

    FidsArray.AFSCBFids_val = 0;
    FidsArray.AFSCBFids_len = 0;
    CBArray.AFSCBs_val = 0;
    CBArray.AFSCBs_len = 0;
    StatArray.AFSBulkStats_val = 0;
    StatArray.AFSBulkStats_len = 0;

This is done for any input or output arguments, but only for types we
need to free afterwards (arrays, usually). We do not do this for
simple types, like single flat structs. In the above example, we do
this for the arrays FidsArray, StatArray, and CBArray, but 'Sync' is
not initialized to anything.

If some server RPC handlers never set a value for an output argument,
this means we'll send uninitialized stack memory to our peer.
Currently this can happen in, for example,
MRXSTATS_RetrieveProcessRPCStats if 'rxi_monitor_processStats' is
unset (specifically, the 'clock_sec' and 'clock_usec' arguments are
never set when rx_enableProcessRPCStats() has not been called).

To make sure we cannot send uninitialized data to our peer, change
rxgen to instead 'memset(&arg, 0, sizeof(arg));' for every single
parameter. Using memset in this way just makes this a little simpler
inside rxgen, since all we need to do this is the name of the
argument.

With this commit, the rxgen-generated code for the above example now
looks like this:

    AFSCBFids FidsArray;
    AFSBulkStats StatArray;
    AFSCBs CBArray;
    AFSVolSync Sync;

    memset(&FidsArray, 0, sizeof(FidsArray));
    memset(&CBArray, 0, sizeof(CBArray));
    memset(&StatArray, 0, sizeof(StatsArray));
    memset(&Sync, 0, sizeof(Sync));

Reviewed-on: https://gerrit.openafs.org/13914
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 93aee3cf40622993b95bd1af77080a31670c24bb)

Change-Id: I6e19aaea57e545455b65851d1bedade584e482f0
Reviewed-on: https://gerrit.openafs.org/13917
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>

7 months agoOPENAFS-SA-2019-001: Skip server OUT args on error 16/13916/2
Andrew Deason [Thu, 8 Aug 2019 01:50:47 +0000]
OPENAFS-SA-2019-001: Skip server OUT args on error

Currently, part of our server-side RPC argument-handling code that's
generated from rxgen looks like this (for example):

    z_result = SRXAFS_BulkStatus(z_call, &FidsArray, &StatArray, &CBArray, &Sync);
    z_xdrs->x_op = XDR_ENCODE;
    if ((!xdr_AFSBulkStats(z_xdrs, &StatArray))
         || (!xdr_AFSCBs(z_xdrs, &CBArray))
         || (!xdr_AFSVolSync(z_xdrs, &Sync)))
            z_result = RXGEN_SS_MARSHAL;
fail:
    [...]
    return z_result;

When the server routine for implementing the RPC results a non-zero
value into z_result, the call will be aborted. However, before we
abort the call, we still call the xdr_* routines with XDR_ENCODE for
all of our output arguments. If the call has not already been aborted
for other reasons, we'll serialize the output argument data into the
Rx call. If we push more data than can fit in a single Rx packet for
the call, then we'll also send that data to the client. Many server
routines for implementing RPCs do not initialize the memory inside
their output arguments during certain errors, and so the memory may be
leaked to the peer.

To avoid this, just jump to the 'fail' label when a nonzero 'z_result'
is returned. This means we skip sending the output argument data to
the peer, but we still free any argument data that needs freeing, and
record the stats for the call (if needed). This makes the above
example now look like this:

    z_result = SRXAFS_BulkStatus(z_call, &FidsArray, &StatArray, &CBArray, &Sync);
    if (z_result)
        goto fail;
    z_xdrs->x_op = XDR_ENCODE;
    if ((!xdr_AFSBulkStats(z_xdrs, &StatArray))
         || (!xdr_AFSCBs(z_xdrs, &CBArray))
         || (!xdr_AFSVolSync(z_xdrs, &Sync)))
            z_result = RXGEN_SS_MARSHAL;
fail:
    [...]
    return z_result;

Reviewed-on: https://gerrit.openafs.org/13913
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit ea276e83e37e5bd27285a3d639f2158639172786)

Change-Id: I688cbf1a65903bf26a0db033687898f3fb5a54ea
Reviewed-on: https://gerrit.openafs.org/13916
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: Benjamin Kaduk <kaduk@mit.edu>

7 months agoMake OpenAFS 1.8.4 41/13841/2 openafs-stable-1_8_4
Stephan Wiesand [Sun, 8 Sep 2019 11:23:23 +0000]
Make OpenAFS 1.8.4

Update version strings for the 1.8.4 release.

Change-Id: I85102e0be54cf0ff28b706d35dba8e3601447cba
Reviewed-on: https://gerrit.openafs.org/13841
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>
Tested-by: Stephan Wiesand <stephan.wiesand@desy.de>

7 months agoUpdate NEWS for OpenAFS 1.8.4 85/13785/3
Stephan Wiesand [Fri, 16 Aug 2019 15:55:48 +0000]
Update NEWS for OpenAFS 1.8.4

Finalize the 1.8.4 release notes.

Change-Id: Ida85451cdec65b2fe736d66572671293f30a6508
Reviewed-on: https://gerrit.openafs.org/13785
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>
Tested-by: Stephan Wiesand <stephan.wiesand@desy.de>

8 months agoMake OpenAFS 1.8.4pre2 39/13839/3 openafs-stable-1_8_4pre2
Stephan Wiesand [Fri, 6 Sep 2019 14:17:39 +0000]
Make OpenAFS 1.8.4pre2

Update version strings for the second 1.8.4 prerelease.

Change-Id: I83ebb06aaae9484e0773b68f336e6d8752069824
Reviewed-on: https://gerrit.openafs.org/13839
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>
Tested-by: Stephan Wiesand <stephan.wiesand@desy.de>

8 months agoUpdate NEWS for 1.8.4pre2 40/13840/2
Stephan Wiesand [Sun, 8 Sep 2019 11:16:35 +0000]
Update NEWS for 1.8.4pre2

Release notes for the OpenAFS 1.8.4 prerelease 2.

Change-Id: Iad2098f56c9e5a9025761e21f609aa3c1584b4aa
Reviewed-on: https://gerrit.openafs.org/13840
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>
Tested-by: Stephan Wiesand <stephan.wiesand@desy.de>

8 months agoRemove one more automake VERSION reference 90/13790/2
Andrew Deason [Wed, 17 Oct 2018 21:35:36 +0000]
Remove one more automake VERSION reference

The configure summary was still referencing the old automake-specific
VERSION var. Use the autoconf PACKAGE_VERSION var instead, so this
actually shows our version.

Reviewed-on: https://gerrit.openafs.org/13360
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 0f65b40b24599d58cf30bfd47fae83ab54e1416a)

Change-Id: I5bd9399acc6e4c6dd19b94198354b600f35bee15
Reviewed-on: https://gerrit.openafs.org/13790
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

8 months agoRemove automake autoconf vars 89/13789/2
Andrew Deason [Thu, 11 Oct 2018 05:18:17 +0000]
Remove automake autoconf vars

Commit 4706854f (autoconf: updates and cleanup) removed our invocation
of AM_INIT_AUTOMAKE, which defines the output variables PACKAGE and
VERSION. Several files in our build system are still referencing
@PACKAGE@ and @VERSION@, though, leaving them un-substituted. This
most easily is seen as the AFSVersion version string remaining as
"@VERSION@" when the tree is built without git, but it also affects
some packaging in the tree.

Remove references to @VERSION@ and @PACKAGE@, replacing them with
their autoconf equivalents @PACKAGE_VERSION@ and @PACKAGE_TARNAME@.

Reviewed-on: https://gerrit.openafs.org/13357
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 2f2c2ce62aa17ecac3651d64c1168af926f7458b)

Change-Id: If2b98b8930bc687170f53f852417fb9374bf6c60
Reviewed-on: https://gerrit.openafs.org/13789
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

8 months agoLINUX 5.3.0: Use send_sig instead of force_sig 88/13788/2
Cheyenne Wills [Fri, 9 Aug 2019 20:25:03 +0000]
LINUX 5.3.0: Use send_sig instead of force_sig

Linux 5.3.0 commit 3cf5d076fb4d48979f382bc9452765bf8b79e740 "signal
Remove task parameter from force_sig" (part of siginfo-linus branch)
changes the parameters for the Linux kernel function force_sig. See LKML
thread starting at https://lkml.org/lkml/2019/5/22/1351

According to the LKML discussion and the above commit message force_sig
is only safe to deliver a synchronous signal to the current task. To
send a signal to another task, we're supposed to use send_sig instead,
which has been available since at least linux 2.6.12-rc12.

Currently, rx_knet calls force_sig to kill the rxk_ListenerTask.  With
the Linux 5.3.0 kernel, this module fails to compile due to the above
noted changes.

Replace the force_sig call with send_sig.  In order to use send_sig, the
rxk_listener thread must allow SIGKILL and during shutdown (umount)
SIGKILL must be unblocked for the rxk_listener thread.

Note that SIGKILL is initially blocked on rxk_listener and is only
unblocked when shutting down the thread.  Having the signal blocked is
sufficient to prevent unwanted signals from reaching the rxk_listener
thread during normal operation.

Reviewed-on: https://gerrit.openafs.org/13753
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 2b7af1243f46496c0b5973b3fa2a6396243f7613)

Change-Id: I6eb44311fbcc63adb6ebeb85a8e076922befd645
Reviewed-on: https://gerrit.openafs.org/13788
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

8 months agoLINUX 5.3.0: Check for 'recurse' arg in keyring_search 87/13787/2
Cheyenne Wills [Thu, 8 Aug 2019 22:53:13 +0000]
LINUX 5.3.0: Check for 'recurse' arg in keyring_search

Linux 5.3.0 commit dcf49dbc8077e278ddd1bc7298abc781496e8a08 "keys: Add a
'recurse' flag for keyring searches" adds a new parameter to
Linux kernel keyring_search function.

Update the call to keyring_search to include the recurse parameter if
available. Setting the parameter to true (1) maintains the current
search behavior.

Change-Id: I038117d1bccdf619a42132fba7d8d61b3ce3c14b
Reviewed-on: https://gerrit.openafs.org/13752
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-on: https://gerrit.openafs.org/13787
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

9 months agoMake OpenAFS 1.8.4pre1 96/13596/3 openafs-stable-1_8_4pre1
Stephan Wiesand [Fri, 17 May 2019 14:52:21 +0000]
Make OpenAFS 1.8.4pre1

Update version strings for the first 1.8.4 prerelease.

Change-Id: I9aa05d7bce308780a1dc815341856e2291579039
Reviewed-on: https://gerrit.openafs.org/13596
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

9 months agoUpdate NEWS for 1.8.4pre1 43/13643/10
Michael Meffie [Fri, 14 Jun 2019 13:30:14 +0000]
Update NEWS for 1.8.4pre1

Release notes for the OpenAFS 1.8.4 prerelease 1.

Change-Id: I3976919660ff6a0eda82b71806e116f4322fb38a
Reviewed-on: https://gerrit.openafs.org/13643
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

9 months agolibadmin: overlap warning in strcpy with gcc9 24/13724/4
Cheyenne Wills [Fri, 5 Jul 2019 14:23:10 +0000]
libadmin: overlap warning in strcpy with gcc9

GCC 9 with --enable-checking produces a new warning/error in
afs_utilAdmin.c associated with a strcpy with the potential of an
overlap.  The index used is signed which triggers the new warning. The
source and target of the strcpy are contained within the same higher
level structure.

Change the variable 'index' from signed to unsigned to resolve the
warning/error. Change the variable 'total' in the same structure to
unsigned to be consistent with it's usage with 'index'.

Reviewed-on: https://gerrit.openafs.org/13660
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 79dffe29c8a0ec55c4231a18077efdfa7c1edf53)

Change-Id: I19a192ecea86314851e6889274eb030c5caff8cb
Reviewed-on: https://gerrit.openafs.org/13724
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

9 months agoaklog: Avoid misleading AFSCELL message 76/13676/2
Andrew Deason [Tue, 30 Oct 2018 20:41:22 +0000]
aklog: Avoid misleading AFSCELL message

Currently, if the AFSCELL environment variable is set, aklog (and
other libauth-using utilities) print out a message when
afsconf_GetLocalCell is called:

    Note: Operation is performed on cell env.example.com

However, this message is also printed (with the AFSCELL cell) when
aklog is given the -cell command-line argument, even though aklog
actually uses the cell given on the command line. For example:

    $ AFSCELL=env.example.com aklog -cell cli.example.com -d
    Note: Operation is performed on cell env.example.com
    Authenticating to cell cli.example.com (server srv1.example.com).
    [...]

libauth will normally not print the "Operation" message if we're not
using the default cell, but it determines this by checking if someone
called afsconf_GetCellInfo before calling afsconf_GetLocalCell. And
currently, aklog calls afsconf_GetLocalCell before
afsconf_GetCellInfo, so the message gets printed because libauth has
no way of knowing that we're actually using a different cell.

klog gets around this by making an additional ignored call to
afsconf_GetCellInfo before afsconf_GetLocalCell, but we can fix this
in aklog by just changing the order of the calls. So, just call
afsconf_GetCellInfo first; if we're using the local cell, we can just
give a NULL cell parameter, instead of looking up the local cellname
first.

Reviewed-on: https://gerrit.openafs.org/13371
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 877d9d79a32b9e81911cb567f844b11c693229f0)

Change-Id: I67350be8c25fb93975442175a64098123503b40c
Reviewed-on: https://gerrit.openafs.org/13676
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoafs: Avoid non-dir ENOENT errors in afs_lookup 92/13692/2
Andrew Deason [Mon, 25 Mar 2019 21:33:39 +0000]
afs: Avoid non-dir ENOENT errors in afs_lookup

Historically, there have been many subsystems in libafs that can
generate ENOENT errors for a variety of reasons. In addition to the
expected case where we lookup a name that doesn't exist, other
scenarios have caused ENOENT error codes to be generated, such as:
internal inconsistencies, I/O errors, or even abort codes from the
network.

When one of these scenarios cause an ENOENT error code in one of those
situations during afs_lookup() when the target name does actually
exist, it can be confusing to a user, or even result in incorrect
application behavior. On Linux in particular, ENOENT results from a
lookup are cached in negative dcache entries, and so can cause future
lookups for the same name to yield ENOENT errors.

Various commits have tried to avoid this abuse of the ENOENT error
code, such as 2aa4cb04 (afs: Stop abusing ENOENT). But we cannot
prevent receiving ENOENT abort codes from the network, and mistakes in
the future may cause more scenarios incorrectly yielding ENOENTs.

However, in afs_lookup, we do know that legitimate ENOENT errors can
only occur in one situation: when we have a valid directory blob, and the
afs_dir_Lookup() operation itself returns an ENOENT error for the
target name. For all other areas of afs_lookup(), we know that an
ENOENT error is not legitimate, since we may not be sure if the target
name exists or not.

So to proactively avoid incorrect ENOENT results, prevent afs_lookup
from returning ENOENT, except in the specific code path where
afs_dir_Lookup is called.

Reviewed-on: https://gerrit.openafs.org/13537
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 5f48367f2bd5bf1c0e689c79508177b649b9113b)

Change-Id: I2698c26d7b75146d92e1763d49dce135ad66f672
Reviewed-on: https://gerrit.openafs.org/13692
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoauth: make PGetTokens2 work with 3-char cellnames 79/13679/4
Mark Vitale [Thu, 23 May 2019 02:50:00 +0000]
auth: make PGetTokens2 work with 3-char cellnames

PGetTokens2 accepts two different types of input:
- an integer 'iterator' to request the nth token set for a user
- a string cellname to request the user's token set for that cell

Unfortunately, it distinguishes between these by assuming if the input
length is sizeof(afs_int32) (4 bytes), it must be an integer.  This
assumption is incorrect if the cellname is three (3) characters long
plus a nul terminator.

The result is that the cellname string is interpreted as a very large
"n"; the subsequent search for the user's "very-large-nth-token" fails,
making it appear that the user has no valid token for this cell.

Improve on this heuristic by double-checking any putative integer input.
If it is actually a 3-character string, then process the input as a
cellname instead.

Introduced by commit 5ec5ad5dcca84e99e5f55987cc4f787cd482fdde 'New
GetToken pioctl'.

While here, add doxygen comments.

Reviewed-on: https://gerrit.openafs.org/13599
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
(cherry picked from commit b0278994826f6bd1dfebc39f26282b8fbdadf1a0)

Change-Id: Ib64749d65f03fc564b1d987b426832442be8d5bd
Reviewed-on: https://gerrit.openafs.org/13679
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoptserver: Check for -restricted in SPR_Delete 88/13688/2
Andrew Deason [Sat, 3 Nov 2018 06:04:43 +0000]
ptserver: Check for -restricted in SPR_Delete

Currently, all prdb write operations, except for SPR_Delete, will fail
with PRPERM if called by a non-system:administrators caller while
restricted mode is active.  SPR_Delete is missing this check, and so
is not affected by the -restricted option.

Fix this by inserting the same check for -restricted as all other code
paths that check for -restricted.

Reviewed-on: https://gerrit.openafs.org/13374
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 2e556c0f23ae439c804352cf51fcf30878b03c7a)

Change-Id: I9a31cf4e6490aa13dc0c239d2660fc146553ee75
Reviewed-on: https://gerrit.openafs.org/13688
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoptserver: Fix AccessOK -restricted for SYSADMINID 87/13687/2
Andrew Deason [Sat, 3 Nov 2018 05:58:58 +0000]
ptserver: Fix AccessOK -restricted for SYSADMINID

According to the documentation, as well as other code paths that check
for -restricted, the -restricted option does not affect members of
system:administrators. Currently, though, AccessOK only bypasses the
-restricted check if the caller is SYSADMINID itself (i.e. localauth).

Fix AccessOK to only do the -restricted checks if the caller is not in
system:administrators, to match the documentation as well as other
ptserver operations.

Reviewed-on: https://gerrit.openafs.org/13373
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 3a8fa4ecd65d5d743fdc573c9f0f261aee2063b6)

Change-Id: I786830efab229a50a521daf3efc624e949475030
Reviewed-on: https://gerrit.openafs.org/13687
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoptserver: Fix AccessOK -restricted for addToGroup 86/13686/2
Andrew Deason [Tue, 30 Oct 2018 19:29:24 +0000]
ptserver: Fix AccessOK -restricted for addToGroup

The function AccessOK is used by all of ptserver RPC handlers that
need to do an authorization check, and the last two arguments are set
as such:

- When adding a member to a group, 'mem' is PRP_ADD_MEM and 'any' is
  PRP_ADD_ANY

- When removing a member from a group, 'mem' is PRP_REMOVE_MEM and
  'any' is 0

- When modifying an entry (setFieldsEntry) or modifying some global
  database fields, 'mem' and 'any' are both set to 0

- When reading an entry and not modifying it, 'mem' and/or 'any' are
  set to other values (depending on if we're checking membership,
  examining the entry itself, etc)

Commit 93ece98c (ptserver-restricted-mode-20050415) added a check to
AccessOK to make it return false for -restricted mode when we are
adding a member to a group, or when 'mem' and 'any' are both 0. This
didn't catch the case when we are removing a member from a group,
though, when 'mem' is PRP_REMOVE_MEM.

It looks like commit a614a8d9 (ptutils-restricted-accessok-20081025)
tried to fix this by adding a check for PRP_REMOVE_MEM, but it also
required 'any' to be set to 0 for the conditional to succeed. This is
true when removing a member from a group, but when adding a member to
a group, 'any' is PRP_ADD_ANY, and so this check fails.

This means that currently, when restricted mode is turned on,
non-admins can still run addToGroup and setFieldsEntry successfully.

Fix this by checking for PRP_ADD_MEM/PRP_REMOVE_MEM separately from
checking if 'mem'/'any' are set to 0. Break up this conditional into
separate if() statements with comments to try to make the checks
more clear.

Reviewed-on: https://gerrit.openafs.org/13370
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit dfc78d533ef64c8d6daf134e2a0f67c5c16f7369)

Change-Id: I7f53570b42e2700a33dd5e72a31f6f7f8b876e79
Reviewed-on: https://gerrit.openafs.org/13686
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoauth: eliminate pointless retries in ktc_ListTokensEx 78/13678/2
Mark Vitale [Thu, 23 May 2019 03:03:11 +0000]
auth: eliminate pointless retries in ktc_ListTokensEx

ktc_ListTokensEx is an iterator to provide the names of each cell for
which a user has a token set.  It does this by looking for the 1 through
nth token set for a given user.  However, as currently implemented,
it always continues searching up to the 100x safety limit even when
there are no more token sets for the user.

Instead, return immediately when VIOC_GETTOK2 returns EDOM (no more
tokens for this user).

Introduced by commit a86ad262d2a8be36f43ab0885a84dde37ddfc464 'auth: Add
the ktc_ListTokensEx function'.

Reviewed-on: https://gerrit.openafs.org/13598
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 95ae30c30d98a3219fd021e0ed83200c1b6c266f)

Change-Id: Ib0489b78b92861ae25b625b3054caf68f3b0a074
Reviewed-on: https://gerrit.openafs.org/13678
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoviced: correct option parsing for -vlru*, -novbc 80/13680/2
Mark Vitale [Thu, 25 Oct 2018 14:27:41 +0000]
viced: correct option parsing for -vlru*, -novbc

Commit a5effd9f1011aa319fdf432c67aec604053b8656 "viced: Use libcmd for
command line options" modernized the option parsing for (da)fileserver,
but introduced a few errors for the following options:

-vlruthresh <nn>
-vlruinterval <nn>
-vlrumax <nn>
-novbc

Correct the errors.

Reviewed-on: https://gerrit.openafs.org/13365
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit d058acb354cab9856303cc341a1f439e4f7f3454)

Change-Id: I5eceb3cc77cfca5c9ec68efdbaabd069c3690c5a
Reviewed-on: https://gerrit.openafs.org/13680
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoutil: serverLog using memory after free 75/13675/2
Cheyenne Wills [Tue, 25 Jun 2019 16:40:53 +0000]
util: serverLog using memory after free

clang's scan-build detected a "use of memory after it is freed"
condition.

The function OpenLogFile frees the variable ourName before creating a
duplicate of the name passed to it.  However there is a call that uses
ourName as the parameter: OpenLogFile(ourName).  This results in freeing
ourName then doing a strdup of the same memory location.

Test the passed parameter and if it's the same as ourName already skip
the free and strdup.

This bug was introduced in commit
    340ec2f79208ee21c3130c4b1c13995947ce426c
    "util: allocate log filename buffers"

Reviewed-on: https://gerrit.openafs.org/13659
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit f5f59cd8d336b153e2b762bb7afd16e6ab1b1ee2)

Change-Id: Ia1ae1144493fa682595b365324322d8d273c0cbb
Reviewed-on: https://gerrit.openafs.org/13675
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoadd dkms dependency in Red Hat unit file 74/13674/2
khm [Tue, 25 Jun 2019 19:51:21 +0000]
add dkms dependency in Red Hat unit file

Currently, there is no explicit relationship between OpenAFS and dkms.
If dkms needs to rebuild the kernel module, OpenAFS will fail to mount
because modprobe will not load the module.  This change specifies that
OpenAFS should run after dkms if dkms is present.

Reviewed-on: https://gerrit.openafs.org/13654
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Laß <lass@mail.uni-paderborn.de>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit fbe2a03aa69bc19768302685d902a25e4d6e157a)

Change-Id: If08df50128a29642e9cdfc311811b99a7889a158
Reviewed-on: https://gerrit.openafs.org/13674
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: khm <khm@sciops.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agopioctl: limit fruitless token searches 39/13639/2
Mark Vitale [Thu, 23 May 2019 02:52:10 +0000]
pioctl: limit fruitless token searches

getNthCell searches the afs_users table for the nth token set belonging to a
given user.  However, it is impossible for a user to have more than one
token set per cell.  If the caller specifies a number greater than the
total number of cells this cache manager knows about, we know the search
will be fruitless.

Instead, return early in this case, avoiding both the lock and the
search.

Reviewed-on: https://gerrit.openafs.org/13597
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit fc7e1700fe84f623fb9163466d24226df00b1a2c)

Change-Id: Idfda263af173a7ca081fcea3eef0ec4a63e66eda
Reviewed-on: https://gerrit.openafs.org/13639
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoFix static expressions in conditionals 38/13638/2
Pat Riehecky [Fri, 1 Jun 2018 21:33:37 +0000]
Fix static expressions in conditionals

The conditions in these if statements are always true (or always false).
Remove the check in cmdebug.c, as it is unnecessary, and fix the check
in vlclient.c to actually check for a valid voltype. (via cppcheck)

Reviewed-on: https://gerrit.openafs.org/13158
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 5cd5cd9fa8754a5af346fa6a392363b046316c75)

Change-Id: Ie3a2d6bfc99d1b5adf0524afc29dac30b655d04d
Reviewed-on: https://gerrit.openafs.org/13638
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoafs: add a file-level comment to afs_osidnlc.c 50/13650/2
Benjamin Kaduk [Fri, 19 Apr 2019 15:38:24 +0000]
afs: add a file-level comment to afs_osidnlc.c

This file doesn't currently do a great job of telling the reader what
it's used for.  Let's give them a hint, especially for the expansion
of "DNLC".

Reviewed-on: https://gerrit.openafs.org/13557
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 3a5ab19fe04058e002bfea90f8b64fab4676de67)

Change-Id: I41c53e5a37cc2da8fa0578a9c5db476530103981
Reviewed-on: https://gerrit.openafs.org/13650
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

10 months agoafs: remove bogus comment from afs_IsTokenExpired 51/13651/2
Mark Vitale [Thu, 13 Jun 2019 03:44:32 +0000]
afs: remove bogus comment from afs_IsTokenExpired

Remove an incorrect comment, introduced with commit
adf2e6e827c6caf55247c5e63b88775393156ae5 'Unix CM: Generalise token
storage'.

No functional change is incurred by this commit.

Reviewed-on: https://gerrit.openafs.org/13640
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 54c34d32e884a5bfb2352e7c8767d743ef3e4647)

Change-Id: I00f91947f9cd64694b2021e484b33dc9074fee94
Reviewed-on: https://gerrit.openafs.org/13651
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Cheyenne Wills <cwills@sinenomine.net>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

11 months agoUse the ppc64le_linuxXX sysname for ppc64le builds 37/13637/2
Andrew Deason [Fri, 17 May 2019 01:01:17 +0000]
Use the ppc64le_linuxXX sysname for ppc64le builds

Commit 191e18eb (Open ppc64le_linux sysname space) added the
ppc64le_linux26 sysname, but it still must be manually specified when
running on ppc64le. Use the ppc64le_linux26 by default on ppc64le, so
we can compile without needing to specify an explicit sysname.

Reviewed-on: https://gerrit.openafs.org/13593
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Mark Vitale <mvitale@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(cherry picked from commit 4b6a4ff31a4197504bbcf2d4c14c24dee672d40e)

Change-Id: Icf8f8b42c499dc42bf5d637dae5ad3e261e68512
Reviewed-on: https://gerrit.openafs.org/13637
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>