cheets_ContainerMount erroneously targets non-ARC processes |
|||||||||||||||||||||
Issue descriptionThree PFQ bots are red, showing the same error: https://cros-goldeneye.corp.google.com/chromeos/healthmonitoring/buildDetails?buildbucketId=8928072375146140480 https://cros-goldeneye.corp.google.com/chromeos/healthmonitoring/buildDetails?buildbucketId=8928072360685360560 https://cros-goldeneye.corp.google.com/chromeos/healthmonitoring/buildDetails?buildbucketId=8928072365721418000
,
Dec 3
TL;DR: I'll wait a few minutes/hours until I have the report from the other boards. It seems that it is failing in veyron-minnie R73-11335. But I don't know whether it is failing in more boards or not. Current history only reports up util 11334: https://stainless.corp.google.com/search?test=%5echeets%5c_ContainerMount$&first_date=2018-11-05&last_date=2018-12-03&row=model&col=build&view=matrix
,
Dec 3
Mmm... what's interesting is that there is not veyron-minnie R73-11335 build in goldeneye: http://cros-goldeneye/chromeos/console/listBuild?boards=veyron-minnie&milestone=73&chromeOsVersion=&chromeVersion=&startTimeFrom=&startTimeTo=&token=AFuIdJCx40V7lErMnwEnyXyIaAiO%3A1543872437112#%2F
,
Dec 3
Issue 911239 has been merged into this issue.
,
Dec 3
PTAL. Thanks.
,
Dec 3
,
Dec 4
Stainless doesn't show failing test (including veyron-minnie R73-11336): https://stainless.corp.google.com/search?test=%5echeets%5c_ContainerMount$&first_date=2018-11-05&last_date=2018-12-03&row=model&col=build&view=matrix Can we close this bug? (btw, hidehiko@ is OOO for several days).
,
Dec 4
Hm, on the other hand, cheets_ContainerMount fails due to another reasons on pantheon. I filed a buganizer bug for that: https://b.corp.google.com/issues/120455962
,
Dec 4
as mentioned in comment#7 closing it as 11336 is green on stainless. wzang@ feel free to reopen it if you think otherwise. thanks.
,
Dec 4
,
Dec 4
The PFQ is red showing the same error. Escalated this to P0. (Also updated the bug description.) Assigning to risan@ who was the last person modifying cheets_ContainerMount.py. Is it possible to do a bisect? Also cc'ed lhchavez@ who is the author of this error message. Do you have more context on this? Thanks!
,
Dec 4
,
Dec 4
Adding drivefs owners.
,
Dec 4
I expect it was https://chromium-review.googlesource.com/c/chromium/src/+/1263677.
,
Dec 4
CL in #15 is reverted.
,
Dec 4
I don't understand what this test wants. Should DriveFS (and I guess other FUSE-mounted filesystems and maybe avfs too) change to use pivot_root somewhere or should the test change?
,
Dec 5
so what needs to happen is that any container in the system (more precisely: any mount namespace that touches /) needs to use pivot_root to avoid accidentally modifying other mount namespaces. the test is just doing its job of alerting of bad usages of mountns, so it shouldn't need updating.
,
Dec 5
It looks like https://chromium-review.googlesource.com/c/chromiumos/platform2/+/513490/ is responsible for FUSE daemons running in their own mount namespaces so this test would probably complain about ntfs or sshfs mounts too if it ever saw them.
,
Dec 5
,
Dec 5
,
Dec 10
Reassigning to sammc@ to ensure the test is fixed before relanding the CL. (Looks like benchan@ is not working on this?)
,
Dec 11
,
Dec 21
According to man(2) pivot_root: "Some of the more obscure uses of pivot_root() may quickly lead to insanity." It's starting to slowly set in. I better leave before it's too late.
,
Dec 21
,
Dec 21
,
Dec 28
This test is ARC-specific and by all signs it shouldn't assert on non-ARC processes, but apparently it was written with assumption that only ARC uses mount namespaces. It must tighten up the list of processes it verifies.
,
Dec 28
,
Jan 2
,
Jan 3
it was intentionally written to assert on non-ARC processes too because there were cases when two unrelated namespaces that did not properly isolate/pivot_root(2) themselves were trampling over each other and causing leaks, mount errors, and general mayhem. ARC was being affected the most and receiving the blame (since ARC heavily uses namespaces and was the most likely to be affected), hence the test was born. At some point I wanted to promote that test to a non-ARC security test, but I never got around to do so :(
,
Jan 7
,
Jan 11
The following revision refers to this bug: https://chromium.googlesource.com/chromiumos/platform2/+/1cf9f3d5cac352295f7f06f24b163b8ff80c3421 commit 1cf9f3d5cac352295f7f06f24b163b8ff80c3421 Author: Sergei Datsenko <dats@google.com> Date: Fri Jan 11 17:47:58 2019 cros-disks: Call pivot_root for FUSE sandboxes BUG= chromium:911236 TEST=unittests, platform_CrosDisksArchive platform_CrosDisksFilesystem platform_CrosDisksSshfs, and manual (check /proc/.../mounts of the process) Change-Id: I82e78f5e6dcd417900dfbaa14e08f06c0fd2870a Reviewed-on: https://chromium-review.googlesource.com/1392584 Commit-Ready: ChromeOS CL Exonerator Bot <chromiumos-cl-exonerator@appspot.gserviceaccount.com> Tested-by: Sergei Datsenko <dats@chromium.org> Reviewed-by: Anand Mistry <amistry@chromium.org> [modify] https://crrev.com/1cf9f3d5cac352295f7f06f24b163b8ff80c3421/cros-disks/fuse_mounter.h [modify] https://crrev.com/1cf9f3d5cac352295f7f06f24b163b8ff80c3421/cros-disks/fuse_helper.cc [modify] https://crrev.com/1cf9f3d5cac352295f7f06f24b163b8ff80c3421/cros-disks/sshfs_helper.cc [modify] https://crrev.com/1cf9f3d5cac352295f7f06f24b163b8ff80c3421/cros-disks/drivefs_helper.cc [modify] https://crrev.com/1cf9f3d5cac352295f7f06f24b163b8ff80c3421/cros-disks/ntfs_mounter.cc [modify] https://crrev.com/1cf9f3d5cac352295f7f06f24b163b8ff80c3421/cros-disks/exfat_mounter.cc [modify] https://crrev.com/1cf9f3d5cac352295f7f06f24b163b8ff80c3421/cros-disks/fuse_mount_manager_test.cc [modify] https://crrev.com/1cf9f3d5cac352295f7f06f24b163b8ff80c3421/cros-disks/fuse_mounter.cc [modify] https://crrev.com/1cf9f3d5cac352295f7f06f24b163b8ff80c3421/cros-disks/sandboxed_process.cc [modify] https://crrev.com/1cf9f3d5cac352295f7f06f24b163b8ff80c3421/cros-disks/sandboxed_process.h [modify] https://crrev.com/1cf9f3d5cac352295f7f06f24b163b8ff80c3421/cros-disks/cros-disks.conf
,
Jan 13
|
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by wzang@chromium.org
, Dec 3