Hello,
While experimenting with bbr protocol, I manipulated the network conditions by maintaining a high RTT for about one second before abruptly reducing it. Some packets sent during the high RTT phase experienced long delays in reaching the destination, while later packets, benefiting from the lower RTT, arrived earlier. This out-of-order arrival triggered the receiver to generate duplicate acknowledgments (dup ACKs). Due to the low RTT, these dup ACKs quickly reached the sender. Upon receiving three dup ACKs, the sender initiated a fast retransmission for an earlier packet that was not lost but was simply taking longer to arrive. Interestingly, despite the fast-retransmitted packet experienced a lower RTT, the original delayed packet still arrived first. When the receiver received this packet, it sent an ACK for the next packet in sequence. However, upon later receiving the fast-retransmitted packet, an issue arose in its logic for updating the acknowledgment number. As a result, even after the next expected packet was received, the acknowledgment number was not updated correctly. The receiver continued sending dup ACKs, ultimately forcing bbr into the retransmission timeout (RTO) phase.
I generated this issue in linux kernel version 5.15.0-117-generic with Ubuntu 20.04. I attempted to confirm whether the issue persists with the latest Linux kernel. However, I discovered that the behavior of bbr has changed in the most recent kernel version, where it now sends chunks of packets instead of sending them one by one over time. As a result, I was unable to reproduce the specific sequence of events that triggered the bug we identified. Consequently, I could not confirm whether the bug still exists in the latest kernel.
I believe that the issue (if still exists) will have to be resolved in the location net/ipv4/tcp_input.c or something like that. There are so many authors here that I do not know who to CC here. So, sending this email to you. Sorry if this is not the best way to report this issue.
Thanks
Shehab
________________________________________
From: Ahmed, Shehab Sarar
Sent: Saturday, February 1, 2025 1:01 PM
To: stable(a)vger.kernel.org
Cc: regressions(a)lists.linux.dev
Subject: TCP Fast Retransmission Issue
Hello,
While experimenting with bbr protocol, I manipulated the network conditions by maintaining a high RTT for about one second before abruptly reducing it. Some packets sent during the high RTT phase experienced long delays in reaching the destination, while later packets, benefiting from the lower RTT, arrived earlier. This out-of-order arrival triggered the receiver to generate duplicate acknowledgments (dup ACKs). Due to the low RTT, these dup ACKs quickly reached the sender. Upon receiving three dup ACKs, the sender initiated a fast retransmission for an earlier packet that was not lost but was simply taking longer to arrive. Interestingly, despite the fast-retransmitted packet experienced a lower RTT, the original delayed packet still arrived first. When the receiver received this packet, it sent an ACK for the next packet in sequence. However, upon later receiving the fast-retransmitted packet, an issue arose in its logic for updating the acknowledgment number. As a result, even after the next expected packet was received, the acknowledgment number was not updated correctly. The receiver continued sending dup ACKs, ultimately forcing bbr into the retransmission timeout (RTO) phase.
I generated this issue in linux kernel version 5.15.0-117-generic with Ubuntu 20.04. I attempted to confirm whether the issue persists with the latest Linux kernel. However, I discovered that the behavior of bbr has changed in the most recent kernel version, where it now sends chunks of packets instead of sending them one by one over time. As a result, I was unable to reproduce the specific sequence of events that triggered the bug we identified. Consequently, I could not confirm whether the bug still exists in the latest kernel.
I believe that the issue (if still exists) will have to be resolved in the location net/ipv4/tcp_input.c or something like that. There are so many authors here that I do not know who to CC here. So, sending this email to you. Sorry if this is not the best way to report this issue.
Thanks
Shehab
Returning to focus on 6.1, here is the 6.1 set from the corresponding
6.6 set:
https://lore.kernel.org/all/20240208232054.15778-1-catherine.hoang@oracle.c…
Two patches are missing from the original set:
[01/21] MAINTAINERS: add Catherine as xfs maintainer for 6.6.y
6.6.y-only change
[16/21] xfs: fix again select in kconfig XFS_ONLINE_SCRUB_STATS
XFS_ONLINE_SCRUB_STATS didn't show up till 6.6
The auto group was run on 10 configs and no regressions were seen.
This has been ack'd on the xfs-stable mailing list.
Thanks,
Leah
Catherine Hoang (1):
xfs: allow read IO and FICLONE to run concurrently
Cheng Lin (1):
xfs: introduce protection for drop nlink
Christoph Hellwig (4):
xfs: handle nimaps=0 from xfs_bmapi_write in xfs_alloc_file_space
xfs: only remap the written blocks in xfs_reflink_end_cow_extent
xfs: clean up FS_XFLAG_REALTIME handling in xfs_ioctl_setattr_xflags
xfs: respect the stable writes flag on the RT device
Darrick J. Wong (8):
xfs: bump max fsgeom struct version
xfs: hoist freeing of rt data fork extent mappings
xfs: prevent rt growfs when quota is enabled
xfs: rt stubs should return negative errnos when rt disabled
xfs: fix units conversion error in xfs_bmap_del_extent_delay
xfs: make sure maxlen is still congruent with prod when rounding down
xfs: clean up dqblk extraction
xfs: dquot recovery does not validate the recovered dquot
Dave Chinner (1):
xfs: inode recovery does not validate the recovered inode
Leah Rumancik (1):
xfs: up(ic_sema) if flushing data device fails
Long Li (2):
xfs: factor out xfs_defer_pending_abort
xfs: abort intent items when recovery intents fail
Omar Sandoval (1):
xfs: fix internal error from AGFL exhaustion
fs/xfs/libxfs/xfs_alloc.c | 27 ++++++++++++--
fs/xfs/libxfs/xfs_bmap.c | 21 +++--------
fs/xfs/libxfs/xfs_defer.c | 28 +++++++++------
fs/xfs/libxfs/xfs_defer.h | 2 +-
fs/xfs/libxfs/xfs_inode_buf.c | 3 ++
fs/xfs/libxfs/xfs_rtbitmap.c | 33 +++++++++++++++++
fs/xfs/libxfs/xfs_sb.h | 2 +-
fs/xfs/xfs_bmap_util.c | 24 +++++++------
fs/xfs/xfs_dquot.c | 5 +--
fs/xfs/xfs_dquot_item_recover.c | 21 +++++++++--
fs/xfs/xfs_file.c | 63 ++++++++++++++++++++++++++-------
fs/xfs/xfs_inode.c | 24 +++++++++++++
fs/xfs/xfs_inode.h | 17 +++++++++
fs/xfs/xfs_inode_item_recover.c | 14 +++++++-
fs/xfs/xfs_ioctl.c | 30 ++++++++++------
fs/xfs/xfs_iops.c | 7 ++++
fs/xfs/xfs_log.c | 23 ++++++------
fs/xfs/xfs_log_recover.c | 2 +-
fs/xfs/xfs_reflink.c | 5 +++
fs/xfs/xfs_rtalloc.c | 33 +++++++++++++----
fs/xfs/xfs_rtalloc.h | 27 ++++++++------
21 files changed, 310 insertions(+), 101 deletions(-)
--
2.48.1.362.g079036d154-goog
From: Tejun Heo <tj(a)kernel.org>
[ Upstream commit 86e6ca55b83c575ab0f2e105cf08f98e58d3d7af ]
blkcg_unpin_online() walks up the blkcg hierarchy putting the online pin. To
walk up, it uses blkcg_parent(blkcg) but it was calling that after
blkcg_destroy_blkgs(blkcg) which could free the blkcg, leading to the
following UAF:
==================================================================
BUG: KASAN: slab-use-after-free in blkcg_unpin_online+0x15a/0x270
Read of size 8 at addr ffff8881057678c0 by task kworker/9:1/117
CPU: 9 UID: 0 PID: 117 Comm: kworker/9:1 Not tainted 6.13.0-rc1-work-00182-gb8f52214c61a-dirty #48
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS unknown 02/02/2022
Workqueue: cgwb_release cgwb_release_workfn
Call Trace:
<TASK>
dump_stack_lvl+0x27/0x80
print_report+0x151/0x710
kasan_report+0xc0/0x100
blkcg_unpin_online+0x15a/0x270
cgwb_release_workfn+0x194/0x480
process_scheduled_works+0x71b/0xe20
worker_thread+0x82a/0xbd0
kthread+0x242/0x2c0
ret_from_fork+0x33/0x70
ret_from_fork_asm+0x1a/0x30
</TASK>
...
Freed by task 1944:
kasan_save_track+0x2b/0x70
kasan_save_free_info+0x3c/0x50
__kasan_slab_free+0x33/0x50
kfree+0x10c/0x330
css_free_rwork_fn+0xe6/0xb30
process_scheduled_works+0x71b/0xe20
worker_thread+0x82a/0xbd0
kthread+0x242/0x2c0
ret_from_fork+0x33/0x70
ret_from_fork_asm+0x1a/0x30
Note that the UAF is not easy to trigger as the free path is indirected
behind a couple RCU grace periods and a work item execution. I could only
trigger it with artifical msleep() injected in blkcg_unpin_online().
Fix it by reading the parent pointer before destroying the blkcg's blkg's.
Signed-off-by: Tejun Heo <tj(a)kernel.org>
Reported-by: Abagail ren <renzezhongucas(a)gmail.com>
Suggested-by: Linus Torvalds <torvalds(a)linuxfoundation.org>
Fixes: 4308a434e5e0 ("blkcg: don't offline parent blkcg first")
Cc: stable(a)vger.kernel.org # v5.7+
Signed-off-by: Jens Axboe <axboe(a)kernel.dk>
Signed-off-by: Andrea Ciprietti <ciprietti(a)google.com>
---
include/linux/blk-cgroup.h | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/include/linux/blk-cgroup.h b/include/linux/blk-cgroup.h
index 0e6e84db06f6..b89099360a86 100644
--- a/include/linux/blk-cgroup.h
+++ b/include/linux/blk-cgroup.h
@@ -428,10 +428,14 @@ static inline void blkcg_pin_online(struct blkcg *blkcg)
static inline void blkcg_unpin_online(struct blkcg *blkcg)
{
do {
+ struct blkcg *parent;
+
if (!refcount_dec_and_test(&blkcg->online_pin))
break;
+
+ parent = blkcg_parent(blkcg);
blkcg_destroy_blkgs(blkcg);
- blkcg = blkcg_parent(blkcg);
+ blkcg = parent;
} while (blkcg);
}
--
2.48.1.262.g85cc9f2d1e-goog