There are 2 open security issues in trixie.
There are 2 open security issues in bookworm.
commit 66151aa9ad01d69daabb0b338cb2f7b94e948734
Author: Christoph Martin <martin@uni-mainz.de>
Date: Sat Oct 17 00:50:53 2026 +0200
add information about dgit workflow
commit 18141c2d882076c8a9535273151528d794e4e2d6
Author: Christoph Martin <martin@uni-mainz.de>
Date: Sat Oct 17 00:41:39 2026 +0200
convert to dgit-maint-merge workflow
commit b1eb4ffe3ff9eea9c1ba46b7aa044209b27b83f2
Author: Christoph Martin <martin@uni-mainz.de>
Date: Sat Oct 17 00:37:15 2026 +0200
change to source/format 1.0
commit ccc6992a965e8ab64361c6875b8bc0436c214060
Merge: c52a807 a7039db
Author: Christoph Martin <martin@uni-mainz.de>
Date: Sat Oct 17 00:34:54 2026 +0200
Declare fast forward from pre-git-based history
commit c52a807cf543f8724d98002af3793083c4b1cd80
Author: Christoph Martin <martin@uni-mainz.de>
Date: Sat Oct 17 00:34:37 2026 +0200
drop existing quilt patch queue
commit 936d5068f6854f22fdb019029d5728e07132fee8
Author: Moritz Schlarb <schlarbm@uni-mainz.de>
Date: Thu Apr 16 13:25:20 2026 +0200
Fix ganesha-top installation without USE_LEGACY_PYTHON_INSTALL
commit 67541f8a04cd9dc5eac58065f073e9079c881f33
Author: Frank Filz <ffilzlnx@mindspring.com>
Date: Sat Oct 17 00:25:59 2026 +0200
0004-RREADDIR-handle-error-conditions-and-fslocations-better
commit 6eedce6435091dab1ad11454d4cd63772956d5a8
Author: Christoph Martin <chrism@debian.org>
Date: Sat Oct 17 00:24:39 2026 +0200
rgb
commit c56b409b132f5e80e47a62056b905e8159415cd7
Author: Christoph Martin <chrism@debian.org>
Date: Sat Oct 17 00:24:25 2026 +0200
spelling
commit 95241c8d41000f296ecc89f8b2c39394b263f95a
Author: Christoph Martin <chrism@debian.org>
Date: Sat Oct 17 00:24:14 2026 +0200
add-latomic-linking-as-needed
commit a863087d177344189a916019a710459db93b58a7
Author: Christoph Martin <chrism@debian.org>
Date: Sat Oct 17 00:24:01 2026 +0200
CMakeLists-LIB_INSTALL_DIR
commit fc40143a4b5f81e93925ee7cfcce67099d24fe9c
Author: Christoph Martin <chrism@debian.org>
Date: Sat Oct 17 00:23:42 2026 +0200
0001-src-CMakeLists.text
commit 91ff67c6e122de3d7d5178d359b9351a8595deab
Author: Christoph Martin <chrism@debian.org>
Date: Sat Oct 17 00:23:22 2026 +0200
0001a-ntirpc-monitoring
commit b18b78a5685e1801f319345d2c62e33cdd8611ab
Merge: 0d27395 76b37c3
Author: Christoph Martin <martin@uni-mainz.de>
Date: Fri Oct 16 18:57:32 2026 +0200
Make fast forward from tarball-based history
commit 0d27395dbe9ec5f0c21eb765ed00bc834df7c843
Author: Christoph Martin <martin@uni-mainz.de>
Date: Fri Oct 16 18:57:04 2026 +0200
Re-import Debian packaging on top of upstream git
commit 76b37c3f7301b288954a1a7796f4bf3049572731
Author: Christoph Martin <martin@uni-mainz.de>
Date: Fri Oct 9 21:24:35 2026 +0200
remove nfs-ganesha-mount-9p, fixes CVE-2026-105516 CVE-2026-105517 (closes: #1150366)
commit a7039dba7c2a8313691f6adb64e47483767481e4
Author: Moritz Schlarb <schlarbm@uni-mainz.de>
Date: Thu Apr 16 13:25:20 2026 +0200
Fix ganesha-top installation without USE_LEGACY_PYTHON_INSTALL
Gbp-Pq: Name 0008-Fix-ganesha-top-installation.patch
commit 55b79d8df20ad61dfd9012440e3ed116d06bbacf
Author: Christoph Martin <chrism@debian.org>
Date: Fri Oct 9 21:23:34 2026 +0200
RREADDIR: handle error conditions and fslocations better
Origin:
Bug: https://github.com/nfs-ganesha/nfs-ganesha/issues/1355
Last-Update: 2026-06-25
Last-Update: 2026-06-25
Gbp-Pq: Name 0004-RREADDIR-handle-error-conditions-and-fslocations-better.patch
commit 9af651a59347c163cb942ee726c7ed3302bc8c97
Author: Christoph Martin <chrism@debian.org>
Date: Fri Oct 9 21:23:34 2026 +0200
fix RGW API version
Last-Update: 2026-02-27
Last-Update: 2026-02-27
Gbp-Pq: Name rgb.patch
commit ad55f58c98f107dfc5d6cbb6308cec07c4b07306
Author: Christoph Martin <chrism@debian.org>
Date: Fri Oct 9 21:23:34 2026 +0200
correct spell errors
Last-Update: 2025-01-10
Last-Update: 2025-01-10
Gbp-Pq: Name spelling.patch
commit 4b9674ce0892c1064afd97ec1f86eabcddd75b4a
Author: Christoph Martin <chrism@debian.org>
Date: Fri Oct 9 21:23:34 2026 +0200
add atomic linker flag
Last-Update: 2026-03-04
Last-Update: 2026-03-04
Gbp-Pq: Name add-latomic-linking-as-needed.patch
commit 0d3b6e560f7d162685e776ce37f14afaa4538fe1
Author: Christoph Martin <chrism@debian.org>
Date: Fri Oct 9 21:23:34 2026 +0200
disable lib64 install
Last-Update: 2025-01-10
Last-Update: 2025-01-10
Gbp-Pq: Name CMakeLists-LIB_INSTALL_DIR.patch
commit 74c5e800e92ffd9f830b5d39ba1b1e08e8124f59
Author: Christoph Martin <chrism@debian.org>
Date: Fri Oct 9 21:23:34 2026 +0200
use debian install layout
Last-Update: 2025-01-10
Last-Update: 2025-01-10
Gbp-Pq: Name 0001-src-CMakeLists.text.patch
commit 2915214dcbc628f2b3a028d0d992886354d0ec30
Author: Christoph Martin <chrism@debian.org>
Date: Fri Oct 9 21:23:34 2026 +0200
backport monitoring functions
Origin: upstream
Last-Update: 2026-03-04
Last-Update: 2026-03-04
Gbp-Pq: Name 0001a-ntirpc-monitoring.patch
commit 37349127da6ab78f9f6a1ed95500c0999a96dfa9
Merge: 288e55b f31f88b
Author: Christoph Martin <chrism@debian.org>
Date: Fri Oct 9 21:23:34 2026 +0200
nfs-ganesha (9.14-2) unstable; urgency=high
* remove nfs-ganesha-mount-9p, fixes CVE-2026-105516 CVE-2026-105517
(closes: #1150366)
[dgit import unpatched nfs-ganesha 9.14-2]
commit f31f88b4b867f2a9ce56f371c484ed29d399096f
Author: Christoph Martin <chrism@debian.org>
Date: Fri Oct 9 21:23:34 2026 +0200
Import nfs-ganesha_9.14-2.debian.tar.xz
[dgit import tarball nfs-ganesha 9.14-2 nfs-ganesha_9.14-2.debian.tar.xz]
commit 288e55bcb4898ac6955d45a5ec6b523af2af6655
Author: Christoph Martin <chrism@debian.org>
Date: Mon Jul 20 11:42:17 2026 +0200
Import nfs-ganesha_9.14.orig.tar.gz
[dgit import orig nfs-ganesha_9.14.orig.tar.gz]
commit 7af850559957b244e712280dd3590c4cbaedaae8
Author: Frank S. Filz <ffilzlnx@mindspring.com>
Date: Thu May 21 11:46:51 2026 -0700
V9.14
Change-Id: I2be041186c0ebef8f4ddb5a1b63e8cd09cf9681a
Signed-off-by: Frank S. Filz <ffilzlnx@mindspring.com>
commit 9be7e07531e2afb65a369f3fcc3e0fce2c922170
Author: Goutham Pacha Ravi <gouthampravi@gmail.com>
Date: Mon May 18 10:58:27 2026 -0700
Allow /0 CIDR prefix in export client blocks
The lexer regex patterns for CIDR suffixes reject prefix
length 0 because they require the first digit to be [1-9].
This prevents valid CIDR notation like 0.0.0.0/0 and ::/0
(RFC 4632, RFC 4291) from being used in export client
blocks to match all addresses in an address family.
Fix the regex to accept /0 while still disallowing leading
zeros. Also fix undefined behavior in cidr_contains_ip()
for IPv4 with mask=0, where a 32-bit shift of an int is
performed (UB in C). Add test cases for /0 parsing and
containment.
Change-Id: Ibd5364266e7a5afe1b2749f9a5fc2c1dc7478ba3
Fixes: https://github.com/nfs-ganesha/nfs-ganesha/issues/1396
Signed-off-by: Goutham Pacha Ravi <gouthampravi@gmail.com>
commit b765b3e886199f6f8992683f2179ee2957d6557b
Author: Animesh Javali <Animesh.Javali@ibm.com>
Date: Wed Mar 25 09:56:04 2026 -0400
Fixed big parameter passed by value. CID 509127,509133
Coverity flagged that struct sau_attr_reply is being
passed by value in isOffsetChangedByClient()
and hasRecentModificationTime(), leading
to unnecessary copying of a large structure.
This change converts the function parameters
to pass by reference(pointer), avoiding the overhead
of copying while maintaining the same functionality.
Change-Id: Ifca860124a05984b60c97e481aec7d85c90cc84e
Signed-off-by: Animesh Javali <Animesh.Javali@ibm.com>
commit 10c55d5168547d0fb0581f8910a5498ad6532f41
Author: Nishant Puri <npuri@redhat.com>
Date: Wed May 6 06:55:13 2026 -0400
log: skip conditional log check on isLevel when no targets configured
isLevel() is on the hot path, called for every log statement.
The call to is_op_context_conditional_flag_set() is always a no-op
when no exports or clients have conditional logging configured,
but still incurs full function call overhead on every invocation.
Introduce a bool conditional_logging_configured. Guard the
is_op_context_conditional_flag_set() call in isLevel() with
unlikely(conditional_logging_configured). In production,
where no conditional logging is active, the branch predictor
skips the call entirely at zero overhead.
Change-Id: Ia91dfedb143bcdf13c2cfd6cb58b8c2efc8b04ed
Signed-off-by: Nishant Puri <npuri@redhat.com>
commit d126f70789ce1c9f80212f9a62fcf0d1577badf4
Author: Suhas Athani <Suhas.Athani@ibm.com>
Date: Tue Apr 28 07:44:11 2026 -0400
Add FullDebug logging for delegation deny decisions
Add LogFullDebug instrumentation to should_we_grant_deleg() for
paths where delegation is not granted.
Add LogFullDebug instrumentation to deleg_supported() for
unsupported delegation checks (global config, object type, FSAL
capability, and export permission checks).
Keep existing decision logic unchanged; logs were added only to
improve observability/troubleshooting.
Change-Id: I9f2ba2401be8a3d73faa66a7c86e7084a28e032e
Signed-off-by: Suhas Athani <Suhas.Athani@ibm.com>
commit ed2c761cdc80ef7ae719b58d7d45b384eafc7230
Author: Sachin Punadikar <sachin.punadikar@ibm.com>
Date: Mon May 18 02:43:10 2026 -0400
RECOVERY: Fix status message
Change-Id: Ic1cbbc0558cc67e63b3d608adf427dda364d7497
Signed-off-by: Sachin Punadikar <sachin.punadikar@ibm.com>
commit 48776241f79bfc7df9e4cb089083bb0110915d29
Author: Sachin Punadikar <sachin.punadikar@ibm.com>
Date: Mon Apr 27 13:47:33 2026 -0400
Recovery: Handle zero clients condition in grace
In a clustered grace setup, there are chances that a Ganesha having 0
clients connected may come out of grace immediately and other Ganesha
which has started grace will fail to detect that other Ganesha has
enforced grace. Need to handle this situation and stop the Ganesha from
quickly ending grace period.
Change-Id: I9f255edad99d322e819621f7ecd8a304fcd46968
Signed-off-by: Sachin Punadikar <sachin.punadikar@ibm.com>
commit 810421b5c3857ef24cacea626b49b5778fa3f69e
Author: Lior Suliman <liorsu@gmail.com>
Date: Tue Apr 21 11:55:17 2026 +0000
Clang-formatted the code base
Change-Id: Ic991c47412d304c730ba539c6528bc6e9ca004b7
Signed-off-by: Lior Suliman <liorsu@gmail.com>
commit f3379326403e16a5f67e897b275569d6eedd2773
Author: Prabhu Murugesan <Prabhu.Murugesan1@ibm.com>
Date: Wed Apr 8 00:34:35 2026 +0530
Fix getgrouplist return value handling for NSS and SSSD
The issue: getgrouplist() (NSS) and sss_nss_getgrouplist_timeout() (SSSD)
have different return value semantics:
- NSS: returns -1 on buffer too small, or >=0 (number of groups) on success
- SSSD: returns 0 on success, ERANGE on buffer too small, or errno on error
This caused problems because SSSD's positive errno values (ERANGE=34,
ETIMEDOUT=110) overlap with NSS's success values (number of groups).
Solution: Normalize return values in pwnam_wrappers__getgrouplist():
- NSS: Convert >=0 (success) to 0
- SSSD: Convert ERANGE (buffer too small) to -1
- Result: -1 (retry), 0 (success), >0 (error)
Update my_getgrouplist_alloc() to handle normalized values:
- ret == -1 and errno ERANGE: Buffer too small, retry with correct size
- ret == -1: Failure
- ret >= 0: Success, continue
Change-Id: I1ef2944172054e5051305049fe33192724ea7e51
Signed-off-by: Lior Suliman <liorsu@gmail.com>
commit 88701318b36aaac1d0dd0f86eff16a875f9fad84
Author: Lior Suliman <liorsu@gmail.com>
Date: Thu Apr 30 13:36:46 2026 +0000
log: Add a compile time assert for default_log_levels
Change-Id: I3d4c34b5f16805efe5abe9936d81a74c3f77a179
Signed-off-by: Lior Suliman <liorsu@gmail.com>
commit 73523de81c28963ff5789e10d2a6000f3367daf0
Author: Lior Suliman <liorsu@gmail.com>
Date: Sun Apr 19 14:37:28 2026 +0000
idmapper: Implement RFC 5661 compliant numeric ID parsing
This patch introduces string_to_numeric_uid_gid to strictly enforce
RFC 5661, Section 5.9 requirements for numeric owners and groups.
The new helper ensures:
No leading zeros (except for the value "0").
No non-numeric characters (e.g., '+' or '-').
Proper 32-bit range validation with error logging.
The atless2id and pwentname2id functions were updated to use this
safer parsing logic. Additionally, name2id now bypasses cache
lookups and defaults to numeric parsing when ID mapping is disabled.
Change-Id: I5f8cd6f9d56041931e54e335d0b03283d1015aca
Signed-off-by: Lior Suliman <liorsu@gmail.com>
commit 13e4120da82df5629e128821c7b138503773b901
Author: Lior Suliman <liorsu@gmail.com>
Date: Sun Apr 19 12:40:46 2026 +0000
network: unify ipv4 and ipv6-mapped address handling
Fix sockaddr comparison and hashing to treat native IPv4 addresses
and their IPv6-mapped counterparts as identical logical endpoints.
Previously, comparing native IPv4 and IPv6-mapped addresses could
yield inconsistent ordering, breaking sorted structures like AVL trees.
The comparison logic now normalizes both addresses to a consistent
format before evaluation to ensure a strict weak ordering.
Additionally, update the hashing function to return identical keys
for IPv4 and equivalent mapped IPv6 addresses, preventing duplicate
entries in hash-based data structures.
Change-Id: I90966c115da059087da5cdeb4ed7b5c1be45dd54
Signed-off-by: Lior Suliman <liorsu@gmail.com>
commit 6fda0a4d145c3e9e2bb98df12cac1eeb009eed0f
Author: Frank S. Filz <ffilzlnx@mindspring.com>
Date: Fri Nov 21 11:25:08 2025 -0800
Make FSAL readlink use utf8string instead of gsh_buffdesc and eliminate gsh_strldup
gsh_strldup was used in several places to make copies of a string in
either a gsh_buffdesc or a utf8string which use different types for the
length. Also the length in gsh_buffdesc is the length of the buffer so
it is string length +1. The length was not being consistently set.
Changing the FSAL readlink method to take a utf8string instead of a
gsh_buffdesc allows for a clear definitions since we assure that
internally all utf8strings are NUL terminated with the length just
being a conveniemce.
Introduced a new function that copies a character buffer (that need not
be NUL terminated) into a utf8string, making a copy of the buffer,
adding a NUL termination, and setting the utf8string_len properly.
This allows elimination of all uses of the gsh_strldup macro that
doesn't convert to an inline function well (due the the unknown type
of the length variable) as part of a subsequent patch to make more
header files usable with C++.
Change-Id: Ifd5a250ba500f005d4345cac0fd1defaa14f8485
Signed-off-by: Frank S. Filz <ffilzlnx@mindspring.com>
commit 0a7276c86613762753d212b59a36156f1dbfe5a7
Author: Frank S. Filz <ffilzlnx@mindspring.com>
Date: Thu Feb 5 13:58:16 2026 -0800
Remove LogAlways
It was only for use by test programs that have all been removed.
Change-Id: Iaf29234e6a02d2f64f51d55e4d893fcd8cec041e
Signed-off-by: Frank S. Filz <ffilzlnx@mindspring.com>
commit 30c392cd9a01f1c6036083679bea7522812fc000
Author: Frank S. Filz <ffilzlnx@mindspring.com>
Date: Fri Apr 24 15:23:03 2026 -0700
V9.13
Change-Id: I998a30a164cc9742bb72f43f8ca406e4cef2ab64
Signed-off-by: Frank S. Filz <ffilzlnx@mindspring.com>
commit 966184978d3f5804ab1856713f0ff83d9656496a
Author: Sreedhar Agraharam <sragraha@redhat.com>
Date: Tue Mar 17 02:20:35 2026 -0400
Enhance header files to support both C and C++ compilation
This change standardizes the header interface, making it
robust for use of both C and C++ source files, which is
critical for prometheus .cc files usability.
Change-Id: Id8355941660d40654dc3a7d2f5293d2570996f8f
Signed-off-by: Sreedhar Agraharam <sragraha@redhat.com>
commit fbb426575e8d5ff10159afe8160a5013b005393d
Author: Yogendra Charya Tangalapally <Yogendra.Tangalapally@ibm.com>
Date: Thu Apr 23 03:58:18 2026 -0700
removed ARGS from EXECUTE_PROCESS command for _GIT_DESCRIBE
When moved from EXEC_PROGRAM to EXECUTE_PROCESS, ARGS was not removed
which resulting in ARGS instead of actual ganesha version in _GIT_DESCRIBE
test-31# ganesha_mgr show version | grep Describe
Git Describe = ARGS
Change-Id: I040b31772f7f08214c3c371f1bb14286826d994e
Signed-off-by: Yogendra Charya Tangalapally <Yogendra.Tangalapally@ibm.com>
commit 32335399f4a7aba561273d9533f524cc88630b98
Author: Sachin Punadikar <sachin.punadikar@ibm.com>
Date: Wed Apr 22 12:16:19 2026 -0400
RECOVERY: Handle errors while updating grace object
In a clustered grace setup, grace state and epoch values of current and
recovery are stored in rados object.
Added error handling when such write is failing.
Aditionally added debug message after grace object is modified with
updated values of current and recovery epoch.
Change-Id: I4c1c3c44923f0c5b15cbd3803f9ab85c682753c4
Signed-off-by: Sachin Punadikar <sachin.punadikar@ibm.com>
commit 4682c1cec3f654747aa354d78b67315454a48f0a
Author: Deeraj.Patil <Deeraj.Patil@ibm.com>
Date: Thu Apr 23 09:55:57 2026 -0400
Code alignment, cosmetic changes.
Change-Id: I797d006e020cf250ed36a99012a950c817d41c00
Signed-off-by: Deeraj.Patil <Deeraj.Patil@ibm.com>
commit 5a75aeccf59ddd48df159ac382e411e7fd898198
Author: Martin Schwenke <mschwenke@ddn.com>
Date: Tue Apr 14 12:27:24 2026 +1000
VFS: Skip dangling directory entries during readdir
After deciding this is a sane thing to do, it turns out there is a
similar check in FSAL_GPFS. It is more ambitious, also matching
ERR_FSAL_STALE and ERR_FSAL_XDEV. However, it is hard to see how the
either case applies here, so be conservative.
Reset status to "no error" to avoid returning ERR_FSAL_NOENT if is the
last directory entry. The FSAL_GPFS code doesn't need to do this
because it uses 2 separate fsal_status_t variables. This case seems
rare enough that we can ignore the cost of resetting status.
Signed-off-by: Martin Schwenke <mschwenke@ddn.com>
Signed-off-by: Peter Schwenke <pschwenke@ddn.com>
Change-Id: I2a8cd0abbb60547eaa71bf4c881a360d1cf11a58
commit f8529c03fcb046cfd17fccce26c0efbdadf09b1e
Author: Peter Schwenke <pschwenke@ddn.com>
Date: Tue Apr 21 16:15:17 2026 +1000
MDCACHE: Skip stale directory entry
This fixes a race condition where a directory entry is found but
when during the ops->getattrs2 call we find that the file has been
deleted.
In the VFS case, this is during vfs_getattr2 it fails when it tries
to lookup the file.
The positioning of the skip needs to be after the callback.
attrs hasn't been filled so, in this case, fsal_release_attrs doesn't
need to be called. However, a skip might be added for some other
reason at a later date.
I do not understand the significance of whence_is_name, so that
should be looked at during the review. It is important, that we
go to check for eod. That is the conditining for exiting the
loop and, maybe, this entry had the eod flag.
Change-Id: I68a22210f69bf6cf88beb199b23dbb1af021c965
Signed-off-by: Peter Schwenke <pschwenke@ddn.com>
commit ca933622f2ce042c9ba9dd5ee13e2988d6630101
Author: Peter Schwenke <pschwenke@ddn.com>
Date: Tue Dec 30 21:01:37 2025 +1100
FSAL: Use Filesystem_Id in NFS file handles
If Filesystem_id is specified in an EXPORT, use it for
the filesystem fsid. This allows multiple directories
to be exported from the same underlying filesystem.
Add config option nfs_param.core_param.fsid_override to
enable the above behaviour.
I have had this patch for a while. I'll submit it
for consideration.
Change-Id: I8505600a77e56fd4cb8214b251bad3a782a98695
Signed-off-by: Peter Schwenke <pschwenke@ddn.com>
commit 63ccdbd5f2b98701e4214cba29cc20919b77238f
Author: Lior Suliman <liorsu@gmail.com>
Date: Tue Apr 21 10:55:26 2026 +0000
log: enforce full coverage of log components at compile time
Remove explicit sizing from the LogComponents and
default_conditional_log_levels arrays, allowing the compiler to
determine their size from the initializer lists.
To maintain safety, this patch introduces a CT_ASSERT macro and
adds compile-time checks to verify that both arrays contain exactly
COMPONENT_COUNT elements. This ensures that any future additions to
the log components will require corresponding updates to these arrays,
preventing silent omissions.
Change-Id: I1a36fa7641f18fa859e011cc81396f64b8c17168
Signed-off-by: Lior Suliman <liorsu@gmail.com>
commit 02dcf884dc64f1bfc04e26408c079036fb064d0c
Author: Lior Suliman <liorsu@gmail.com>
Date: Thu Apr 16 13:44:21 2026 +0000
SAL: Add "none" recovery backend
This commit adds a new recovery backend option called "none".
When selected, it uses the default (empty) recovery backend,
effectively disabling recovery persistence. This is useful for
environments where recovery is handled externally or not required.
Change-Id: Ia0038928bc5472e7abbd8b4b89a523a07de8274a
Signed-off-by: Lior Suliman <liorsu@gmail.com>
commit 3327ac533c6997525cd58ef1d6a27556be97dcd9
Author: Ofir Vainshtein <ofirvins@google.com>
Date: Sun Apr 19 06:36:54 2026 +0000
Add a rolling log index to Ganesha logs.
Introduces a new log format option, LOG_INDEX, which adds a monotonically increasing index to each log line. The index is a 5-digit number that wraps around at 100000. This helps in tracking the order of log messages. The super_ut configuration is updated to enable this new option.
note because index is taken before write and not under exclusive lock with it. logs are not guaranteed to appear in order in the file
Change-Id: Ieb251326238810d9c3251b36a0d40c2d4844d52d
Signed-off-by: Ofir Vainshtein <ofirvins@google.com>
commit 14308c21cd18b2147f0e53c27d7a5393ac27f54e
Author: Frank S. Filz <ffilzlnx@mindspring.com>
Date: Fri Apr 17 17:13:30 2026 -0700
V9.12
Change-Id: Ie371aa2a49807cdb0523d5e40d383a33be7aba1f
Signed-off-by: Frank S. Filz <ffilzlnx@mindspring.com>
commit 32bcd147f5c3718651eb86c6a521395131ddd89e
Author: Animesh Javali <Animesh.Javali@ibm.com>
Date: Thu Apr 16 13:42:36 2026 -0400
IBMCEPH-14009: Deleg type 5 support for ALLOCATE op
With NFSv4.2, when the server grants
"OPEN_DELEGATE_WRITE_ATTRS_DELEG", workloads that trigger
"ALLOCATE" via nfs4_op_allocate hit stateid validation
failures in the server.
On the client side, this leads to repeated attempts to
"TEST_STATEID" via nfs4_op_test_stateid operations,
causing the I/O to stall.
Added code to include OPEN_DELEGATE_WRITE_ATTRS_DELEG
along with regular write delegation.
Change-Id: Ibde44b8e99fc4b9557d21470452c0a05ef333097
Signed-off-by: Animesh Javali <Animesh.Javali@ibm.com>
commit 6d79ab941bdc2b26299ad700df7c9df091e661ac
Author: Sachin Punadikar <sachin.punadikar@ibm.com>
Date: Wed Apr 8 06:07:38 2026 -0400
Recovery - Additional logging for better debugging
Changed log level, added some new logs for better debuuging of NFS
Ganesha reclaim related issues.
Change-Id: I4b495973ae5b76024d68ca2038e51c4a320c0b0a
Signed-off-by: Sachin Punadikar <sachin.punadikar@ibm.com>
commit c216957207a87d2c290d78b03b3ea11f34516ebf
Author: Frank S. Filz <ffilzlnx@mindspring.com>
Date: Fri Apr 17 15:25:54 2026 -0700
config_samples/config.txt - fixup formatting
Change-Id: I968cf46a274e9db49553fbba806375a0277e8eca
Signed-off-by: Frank S. Filz <ffilzlnx@mindspring.com>
commit acbaa7c48560944b7f718407c69e1430de5ce567
Author: Ofir Vainshtein <ofirvins@google.com>
Date: Sun Apr 12 06:13:29 2026 +0000
Add log rotation functionality to Ganesha.
This change introduces the ability to rotate Ganesha log files based on size and/or time. A new "Rotate" block can be added within the "LOG" section of the configuration. The options are:
- size_kb: Rotate when the log file size exceeds this value in KB.
- time_sec: Rotate when the time since the last rotation exceeds this value in seconds.
When rotation occurs, the current log file is renamed to <log_file_path>.old.
Change-Id: I08f410b615874eadcd1a740ad79d87b8c6aae772
Signed-off-by: Ofir Vainshtein <ofirvins@google.com>
commit 98886cfa33d4fa9c00f0938212794c6bec9024da
Author: Lior Suliman <liorsu@gmail.com>
Date: Wed Apr 15 15:34:47 2026 +0000
Add missing include
Change-Id: Ide54f4008c7676e3267c6e3e40a44a1ab9cf2b11
Signed-off-by: Lior Suliman <liorsu@gmail.com>
commit 85b8a38d9f61bc17b9cb90ac5f982139caf9b62e
Author: Lior Suliman <liorsu@gmail.com>
Date: Wed Apr 15 15:29:24 2026 +0000
Add missing log component to default conditinal log levels
This can cause a buffer overflow when setting a level for all logs
Change-Id: I709333e5dd6aeaecd6b61560785e48f64290a355
Signed-off-by: Lior Suliman <liorsu@gmail.com>
commit c892d010354c62a6b3b04282a4fc2559b0f882ee
Author: Suhas Athani <Suhas.Athani@ibm.com>
Date: Wed Apr 15 01:23:16 2026 -0400
Avoid same-client lock-path delegation recall in FSAL_UP (fix recall24)
This change fixes recall24 by preventing spurious NFS delegation recall
when a backend upcall arrives during same-client lock/open-mode
transitions.
With Ceph delegation callbacks, lock-driven reopen activity could trigger up_async_delegrecall, leading to delegrecall_one/CB_RECALL even when all active locks belong to the same client that owns the delegation.
Fix:
Added a guard in FSAL_UP (delegrecall_impl_per_state) to skip recall when:
- the file has active locks, and
- all active NFSv4 lock owners map to the same clientid as the delegation
owner.
In that case, delegation state is restored to DELEG_GRANTED and CB_RECALL is not issued.
This test case is from nfstest delegation suite -
https://github.com/kofemann/nfstest/blob/master/test/nfstest_delegation
Change-Id: I9760d5db1d7cc433501df77abb9048ab8e1b4d97
Signed-off-by: Suhas Athani <Suhas.Athani@ibm.com>
commit 065784ed42b40159db71cc820c8b19cf03731df4
Author: Frank S. Filz <ffilzlnx@mindspring.com>
Date: Fri Apr 10 15:18:38 2026 -0700
V9.11
Change-Id: I349d34c5066ee153a2b4ee06b3de5d3bda80de60
Signed-off-by: Frank S. Filz <ffilzlnx@mindspring.com>
commit 882fd26c20c5ea6fc93063d2551bbab19a270485
Author: Marcus Watts <mwatts@redhat.com>
Date: Sun Mar 22 10:10:59 2026 -0400
CONFIG - fix for unknown block "rados_urls"
The logic in find_unused_blocks() to identify unused configuration
nodes relies on the "found" flag in the node beign set, but but it's
not possible for logic in conf_url_rados.c to set it. I've modified
config_GetBlockNode to unconditionally set this, since there doesn't
seem to be any valid reason to not set this flag right there.
As it turns out, conf_url_rados is the only user of this routine.
Change-Id: I4dd4bd5b3ffa6c44b6d2823885344fa662243519
Signed-off-by: Marcus Watts <mwatts@redhat.com>
Among the 8 debian patches available in version 9.14-2 of the package, we noticed the following issues:
![[bug history graph]](https://qa.debian.org/data/bts/graphs/n/nfs-ganesha.png){kind=link}