When testing CRIU with containers with docker images, we found an easily reproducible kernel crash:
[ 2300.870181] BUG: kernel NULL pointer dereference, address: 0000000000000005 [ 2300.871537] #PF: supervisor read access in kernel mode [ 2300.872577] #PF: error_code(0x0000) - not-present page [ 2300.873595] PGD 800000010a18f067 P4D 800000010a18f067 PUD 10730c067 PMD 0 [ 2300.874900] Oops: Oops: 0000 [#1] SMP PTI [ 2300.875742] CPU: 1 UID: 0 PID: 5419 Comm: criu ve: 0 Kdump: loaded Not tainted 6.12.0-211.39.1.16.6.vz10.x86_64 #1 16.6 PREEMPT(voluntary) [ 2300.878057] Hardware name: Acronis OpenStack Compute/Virtuozzo, BIOS 1.16.1-1.vz9.2 04/01/2014 [ 2300.879684] RIP: 0010:ovl_get_index_name_fh+0x18/0x70 [overlay] [ 2300.880859] Code: ... [ 2300.884128] RSP: 0018:ffffd32582e0f7c0 EFLAGS: 00010246 [ 2300.895807] Call Trace: [ 2300.896397] <TASK> [ 2300.904435] ovl_create_index+0x82/0x260 [overlay] [ 2300.905406] ovl_copy_up_workdir+0x30a/0x330 [overlay] [ 2300.906427] ovl_do_copy_up+0x10d/0x2e0 [overlay] [ 2300.907387] ovl_copy_up_one+0x456/0x720 [overlay] [ 2300.910736] ovl_copy_up_flags+0x81/0x110 [overlay] [ 2300.911740] ovl_create_object+0x51/0x130 [overlay] [ 2300.912719] vfs_mkdir+0x194/0x250 [ 2300.913470] do_mkdirat+0x14c/0x180 [ 2300.914223] __x64_sys_mkdir+0x46/0x70 [ 2300.915019] do_syscall_64+0x7d/0x160 [ 2300.922814] entry_SYSCALL_64_after_hwframe+0x76/0x7e The cause is that overlayfs does some feature checks on superblock creation using capable/ve_capable and it stores current credentials. But ve_capable check result depends on both current creds and current VE. Later, overlayfs switches credentials only, leading to unexpected feature disappearance. 1. Patch 1 removes separate STOPPED state - VE goes directly to DEAD. 2. Patch 2 adds helpers for overriding tasks VE. 3. Patch 3 delays ve cred drop until final destruction of VE. 4. Patch 4 adds VE owner to overlay superblock. 5. Patch 5 fixes fd leak that appeared after rebase. 6. Patch 6 fixes the bug by switching VE together with creds in overlayfs. As a side effect of this series - CT1 overlayfs mounts won't work in CT2 and this is a good thing. There's still a gap - creds may be overrided in other places and we should handle them - a separate VSTOR-155347 task is for it. https://virtuozzo.atlassian.net/browse/VSTOR-143259 - non-CRIU repro https://virtuozzo.atlassian.net/browse/VSTOR-142405 Signed-off-by: Vladimir Riabchun <[email protected]> --- Changes in v8: - One more time updated commit messages to make sashiko happy. - Rebased on hci-8.2. - Returned ve_name print in patch 2 - now we free name in the very end. - In patch 6 did noted other cred usages in commit message and updated ovl_lower_positive to return true on failure. - Link to v7: http://lore.virtuozzo.com/[email protected] Changes in v7: - Patch 2: dropped new_ve name printing due to a race with free. - Added new patch to fix rebase-introduced fd leak. - Updated some commit messages. - Link to v6: http://lore.virtuozzo.com/[email protected] Changes in v6: - Fixed various typos spotted by Kostya. - Updated comments in commit messages and code. - Rebased on rh10-6.12.0-211.75.1.24.1.vz10, overlayfs got one more place with creds override ovl_copyfile. - Patch 2: helper always returns -EPERM for non-host switches. - Patch 5: CT -> Host switch is handled by overlayfs now, code updated with Kostya's suggestions. - Link to v5: http://lore.virtuozzo.com/[email protected] Changes in v5: - Patch 2: report non-host VE changes, also treat VEx -> VE0 and VEx -> VEy differently. VEx -> VE0 might be allowed, VEx -> VEy - never. Replaced most WARN_ON with normal warnings or WARN_ON_ONCE. - Patch 3: updated commit name. - Patch 5: allow VEx -> VE0 case. Now creds are switched without VE. - Link to v4: http://lore.virtuozzo.com/[email protected] Changes in v4: - Patch 2: clean non-NULL overrided_ve in fork and report it, Report revert_ve(ERR_PTR) usages and reject them. - Patch 5: changed ovl_override_creds signature so it returns error. Added checks for it in all places to avoid. - Link to v3: http://lore.virtuozzo.com/[email protected] Changes in v3: - Patch 2: added a separate field for ve_override to fix problems Pasha spotted. - Patch 3: new one, delays init_cred destroy so dead CT can still use it. - Link to v2: http://lore.virtuozzo.com/[email protected] Changes in v2: - Patch 1: updated description as Sashiko suggested.. - Patch 2: moved stubs outside CONFIG_X86 section, added task_lock protection in helpers. Verified via lockdep, no splats after vzctl tests, C/R and crash reproducer. - Patch 4: fixed typo in comment. - Link to v1: http://lore.virtuozzo.com/[email protected] --- Vladimir Riabchun (6): ve: Remove STOPPED state ve: Introduce override_ve/revert_ve helpers ve: Defer ve creds drop until ve is completely destroyed ovl: Keep creator VE in overlayfs superblock ovl: Fix leaking fd references in ovl_copyfile ve/overlayfs: Override VE together with creds fs/overlayfs/copy_up.c | 6 ++-- fs/overlayfs/dir.c | 59 ++++++++++++++++++------------ fs/overlayfs/file.c | 74 +++++++++++++++++++++++--------------- fs/overlayfs/inode.c | 94 +++++++++++++++++++++++++++++++----------------- fs/overlayfs/namei.c | 39 +++++++++++++------- fs/overlayfs/overlayfs.h | 10 +++++- fs/overlayfs/ovl_entry.h | 1 + fs/overlayfs/params.c | 3 ++ fs/overlayfs/readdir.c | 36 ++++++++++++------- fs/overlayfs/super.c | 3 ++ fs/overlayfs/util.c | 53 ++++++++++++++++++++++----- fs/overlayfs/xattrs.c | 35 +++++++++++------- include/linux/sched.h | 1 + include/linux/ve.h | 33 +++++++++++++++-- kernel/capability.c | 10 +++--- kernel/fork.c | 2 ++ kernel/ve/ve.c | 10 +++--- net/netlink/af_netlink.c | 2 +- 18 files changed, 325 insertions(+), 146 deletions(-) --- base-commit: d816dd0e4cdc45208ba0c488d295c5b2f943e079 change-id: 20260918-ovl-use-orig-ve-2dc365976202 Best regards, -- Vladimir Riabchun <[email protected]> _______________________________________________ Devel mailing list [email protected] https://lists.openvz.org/mailman/listinfo/devel
