cp: better permissions mode handling with umask awareness by my4ng · Pull Request #10904 · uutils/coreutils

my4ng · 2026-02-13T04:41:21Z

Fix #9750, #10011, #10787, #10862, Supersede #10859.

This PR reworks the permissions mode handling such that its behaviour is in line with GNU's implementation. Specifically:

set umask to !0o077, which means (1) cp can actually create and work with new directories/files which will have 700 permissions mode before their final calculated mode is set. the original umask is kept and passed around for calculation in the case of newly created and non-mode-preserving dir/files. umask is restored to original on exit.
for explicit non-mode-preserving, newly created dir/files, use 777 and 666 modes for calculation respectively.
for existing dir/files, leave existing mode unchanged except when preserving mode.
for newly created dir/files, always set their permission mode to the correct one.
~~added 12 new tests for the combination of existing/not-existing, no-preserve/default/preserve, dir/file~~ one each for directory and file with rstest.
improve performance by eliminating redundant heap allocations (e.g. PathBuf, String), ~25% from benchmark.

github-actions · 2026-02-13T04:55:44Z

GNU testsuite comparison:

GNU test failed: tests/cp/cp-parents. tests/cp/cp-parents is passing on 'main'. Maybe you have to rebase?
Skipping an intermittent issue tests/tty/tty-eof (passes in this run but fails in the 'main' branch)
Congrats! The gnu test tests/factor/t34 is no longer failing!
Congrats! The gnu test tests/printf/printf-surprise is now passing!

ChrisDryden · 2026-02-13T10:17:01Z

Is there any way to split this PR up into something smaller? When its this big it gets pretty hard to review

my4ng · 2026-02-13T10:33:14Z

I'm afraid not. Most changes are actually from the noise of propagating orig_umask, and the only significant changes being build_dir and handle_mode. Without either, it results in a dozen or more failed tests.

There is a GNU test regression: tests/cp/cp-parents, caused by the second to last test in that file due to created parent dir not having attributes set correctly. Will fix soon.

github-actions · 2026-02-13T14:41:12Z

GNU testsuite comparison:

GNU test failed: tests/pr/bounded-memory. tests/pr/bounded-memory is passing on 'main'. Maybe you have to rebase?
Skip an intermittent issue tests/tail/symlink (fails in this run but passes in the 'main' branch)
Congrats! The gnu test tests/factor/t34 is no longer failing!
Note: The gnu test tests/tail/pipe-f is now being skipped but was previously passing.
Congrats! The gnu test tests/printf/printf-surprise is now passing!

github-actions · 2026-02-14T05:38:16Z

GNU testsuite comparison:

GNU test failed: tests/cut/bounded-memory. tests/cut/bounded-memory is passing on 'main'. Maybe you have to rebase?
GNU test failed: tests/misc/io-errors. tests/misc/io-errors is passing on 'main'. Maybe you have to rebase?
GNU test failed: tests/pr/bounded-memory. tests/pr/bounded-memory is passing on 'main'. Maybe you have to rebase?
Skipping an intermittent issue tests/tty/tty-eof (passes in this run but fails in the 'main' branch)
Congrats! The gnu test tests/factor/t10 is no longer failing!

github-actions · 2026-02-21T14:04:38Z

GNU testsuite comparison:

Congrats! The gnu test tests/tail/tail-n0f is now passing!

codspeed · 2026-02-21T14:11:51Z

Merging this PR will improve performance by 50.18%

⚠️

Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 5 improved benchmarks
✅ 304 untouched benchmarks
⏩ 46 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	`BASE`	`HEAD`	Efficiency
⚡	Simulation	`cp_archive_balanced_tree[(5, 4, 10)]`	76.8 ms	62.1 ms	+23.7%
⚡	Simulation	`cp_recursive_wide_tree[(6000, 800)]`	187.8 ms	156.1 ms	+20.28%
⚡	Simulation	`cp_preserve_metadata[(5, 4, 10)]`	73.7 ms	59.2 ms	+24.43%
⚡	Simulation	`cp_recursive_balanced_tree[(5, 4, 10)]`	71.4 ms	56.6 ms	+26.06%
⚡	Simulation	`cp_recursive_deep_tree[(120, 4)]`	13.1 ms	8.7 ms	+50.18%

_{Comparing my4ng:cp_mode (8d082ea) with main (b7e967c)}

46 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

my4ng · 2026-02-21T14:32:13Z

There are some residual problems with copy_file that I want to fix. In particular, currently it is setting the permissions (unnecessarily) twice, but the change appears to be more complex than I anticipated. Might split that into a separate PR just for clarity since this one is working and big enough already.

github-actions · 2026-03-17T02:33:45Z

GNU testsuite comparison:

Skipping an intermittent issue tests/pr/bounded-memory (passes in this run but fails in the 'main' branch)
Note: The gnu test tests/cut/bounded-memory is now being skipped but was previously passing.
Congrats! The gnu test tests/expand/bounded-memory is now passing!

my4ng · 2026-03-17T23:02:38Z

@ChrisDryden I have condensed the tests into two instead of the original 12, so it should be a lot more workable now.

github-actions · 2026-03-17T23:18:29Z

GNU testsuite comparison:

Congrats! The gnu test tests/misc/io-errors is no longer failing!

hphilm · 2026-04-14T08:45:07Z

Seems with this even applied on top of 0.8.0 we have issues in AerynOS with cp -R and cp -a still using the source directory permissions of 770 even umask 0022 is given to get 755 normally. Here our test results:

# bash test.sh

Set my umask to 0022
What's my umask?
0022

whoami?
root

Which cp version are we using?
cp (uutils coreutils) 0.8.0

>> File permissions test:

Create a 0775 | rwxrwxr-x file:
mode of 'file.0775' changed from 0644 (rw-r--r--) to 0775 (rwxrwxr-x)
-rwxrwxr-x 1 root root 0 Apr 14 08:27 file.0775

Assumption: cp -Rv respects 0022 umask whether or not the user is root
'file.0775' -> 'file.0755'
Test assumption:
-rwxr-xr-x 1 root root 0 Apr 14 08:27 file.0755

>> Directory permissions test:

Create a 0775 | rwxrwxr-x directory:
mode of 'dir.0775' changed from 0755 (rwxr-xr-x) to 0775 (rwxrwxr-x)
drwxrwxr-x 1 root root 0 Apr 14 08:27 dir.0775

Assumption: cp -Rv respects 0022 umask whether or not the user is root
'dir.0775' -> 'dir.0755/'
Test assumption:
drwxrwxr-x 1 root root 0 Apr 14 08:27 dir.0755

Script we used to test it:

#!/usr/bin/env bash
echo -e "\nSet my umask to 0022"
umask 0022
echo -e "What's my umask?"
umask
echo -e "\nwhoami?"
whoami
echo -e "\nWhich cp version are we using?"
cp --version

echo -e "\n>> File permissions test:"
echo -e "\nCreate a 0775 | rwxrwxr-x file:"
touch file.0775 && chmod -c u=rwx,g=rwx,o=rx file.0775 && ls -l file.0775
echo -e "\nAssumption: cp -Rv respects 0022 umask whether or not the user is root"
cp -Rfv file.0775 file.0755
echo "Test assumption:"
ls -l file.0755

echo -e "\n>> Directory permissions test:"
echo -e "\nCreate a 0775 | rwxrwxr-x directory:"
[[ -d dir.0775 ]] && rm -rf dir.0775
mkdir dir.0775 && chmod -c u=rwx,g=rwx,o=rx dir.0775 && ls -ld dir.0775
echo -e "\nAssumption: cp -Rv respects 0022 umask whether or not the user is root"
[[ -d dir.0755 ]] && rm -rf dir.0755
cp -Rfv dir.0775 dir.0755
echo "Test assumption:"
ls -ld dir.0755

Using this patch on top of 0.8.0 without your patches restores the 0.6.0 behaviour ...

root@boulder:~/v0.76.0.tar.gz/a
# bash test.sh

Set my umask to 0022
What's my umask?
0022

whoami?
root

Which cp version are we using?
cp (uutils coreutils) 0.8.0

>> File permissions test:

Create a 0775 | rwxrwxr-x file:
mode of 'file.0775' changed from 0644 (rw-r--r--) to 0775 (rwxrwxr-x)
-rwxrwxr-x 1 root root 0 Apr 14 07:49 file.0775

Assumption: cp -Rv respects 0022 umask whether or not the user is root
'file.0775' -> 'file.0755'
Test assumption:
-rwxr-xr-x 1 root root 0 Apr 14 07:49 file.0755

>> Directory permissions test:

Create a 0775 | rwxrwxr-x directory:
mode of 'dir.0775' changed from 0755 (rwxr-xr-x) to 0775 (rwxrwxr-x)
drwxrwxr-x 1 root root 0 Apr 14 07:49 dir.0775

Assumption: cp -Rv respects 0022 umask whether or not the user is root
'dir.0775' -> 'dir.0755/'
Test assumption:
drwxr-xr-x 1 root root 0 Apr 14 07:49 dir.0755

ermo · 2026-04-15T20:46:35Z

As a follow-up to the above post, it would appear that the basic takeaway from the above test script that I wrote for the AerynOS uutils-coreutils cp investigation (and also tested w/ GNU coreutils cp) is that IFF --preserve is NOT set, THEN the new files or directories being created by cp and cp -R should respect the umask.

IF we assume that source permissions & !umask == max target permissions THEN:

umask = 0000: source permissions & !0000 => max target perms: u=7&7=7=rwx, g=7&7=7=rwx, o=7&7=7=rwx
umask = 0002: source permissions & !0002 => max target perms: u=7&7=7=rwx, g=7&7=7=rwx, o=7&5=5=r-x
umask = 0022: source permissions & !0022 => max target perms: u=7&7=7=rwx, g=7&5=5=r-x, o=7&5=5=r-x
umask = 0222: source permissions & !0222 => max target perms: u=7&5=5=r-x, g=7&5=5=r-x, o=7&5=5=r-x

Effectively, if the source permissions do not reach the umask bitfilter "cap", the target permissions are the same as the source permissions.

There may be different implications for the setuid/setgid/sticky bits. Those have not been accounted for here or in the above script.

hphilm · 2026-04-16T01:38:17Z

I think that proofing with my patch on the current 0.8.0 code, preserve is set on new created folders regardless and therefore umask ignored completely...

Fix CI-failing checks.

my4ng · 2026-04-16T06:43:10Z

@hphilm I have rebased my PR to the main branch, and ran your test script and it behaves as expected:

Set my umask to 0022
What's my umask?
0022

whoami?
root

Which cp version are we using?
cp (uutils coreutils) 0.8.0

>> File permissions test:

Create a 0775 | rwxrwxr-x file:
mode of 'file.0775' changed from 0644 (rw-r--r--) to 0775 (rwxrwxr-x)
-rwxrwxr-x. 1 root root 0 Apr 16 16:38 file.0775

Assumption: cp -Rv respects 0022 umask whether or not the user is root
'file.0775' -> 'file.0755'
Test assumption:
-rwxr-xr-x. 1 root root 0 Apr 16 16:38 file.0755

>> Directory permissions test:

Create a 0775 | rwxrwxr-x directory:
mode of 'dir.0775' changed from 0755 (rwxr-xr-x) to 0775 (rwxrwxr-x)
drwxrwxr-x. 1 root root 0 Apr 16 16:38 dir.0775

Assumption: cp -Rv respects 0022 umask whether or not the user is root
'dir.0775' -> 'dir.0755/'
Test assumption:
drwxr-xr-x. 1 root root 0 Apr 16 16:38 dir.0755

All the directory/file tests still pass. Are you using the latest and complete version of this PR?

github-actions · 2026-04-16T06:52:01Z

GNU testsuite comparison:

Note: The gnu test tests/cut/cut-huge-range is now being skipped but was previously passing.
Note: The gnu test tests/rm/many-dir-entries-vs-OOM is now being skipped but was previously passing.

hphilm · 2026-04-16T08:22:55Z

I just pulled the PR again and compiled it against 4e68e67. However, I still see the "wrong" permissions on AerynOS:

root@boulder:~/4e68e67.tar.gz
# bash test.sh

Set my umask to 0022
What's my umask?
0022

whoami?
root

Which cp version are we using?
cp (uutils coreutils) 0.8.0

>> File permissions test:

Create a 0775 | rwxrwxr-x file:
mode of 'file.0775' changed from 0644 (rw-r--r--) to 0775 (rwxrwxr-x)
-rwxrwxr-x 1 root root 0 Apr 16 08:16 file.0775

Assumption: cp -Rv respects 0022 umask whether or not the user is root
'file.0775' -> 'file.0755'
Test assumption:
-rwxr-xr-x 1 root root 0 Apr 16 08:16 file.0755

>> Directory permissions test:

Create a 0775 | rwxrwxr-x directory:
mode of 'dir.0775' changed from 0755 (rwxr-xr-x) to 0775 (rwxrwxr-x)
drwxrwxr-x 1 root root 0 Apr 16 08:16 dir.0775

Assumption: cp -Rv respects 0022 umask whether or not the user is root
'dir.0775' -> 'dir.0755/'
Test assumption:
drwxrwxr-x 1 root root 0 Apr 16 08:16 dir.0755

See my changes here: aerynOS/recipes@3b4ce86
We might need some verbose debug lines to see which permissions are seen and which are set in the end.

my4ng · 2026-04-23T07:08:22Z

@hphilm Could you please provide the strace for the erring directory copy?

hphilm · 2026-05-07T14:59:33Z

Will test with latest main again as 7ba1bf8 might change some we might need.

hphilm · 2026-05-12T06:18:20Z

Your PR needs a rebase on 089db52 or later. I've tested main, but directory copy still fails on AerynOS:

root@boulder:~/x86_64
# bash ./test.sh

Set my umask to 0022
What's my umask?
0022

whoami?
root

Which cp version are we using?
cp (uutils coreutils) 0.8.0

>> File permissions test:

Create a 0775 | rwxrwxr-x file:
-rwxrwxr-x 1 root root 0 May 12 06:14 file.0775

Assumption: cp -Rv respects 0022 umask whether or not the user is root
'file.0775' -> 'file.0755'
Test assumption:
-rwxr-xr-x 1 root root 0 May 12 06:14 file.0755

>> Directory permissions test:

Create a 0775 | rwxrwxr-x directory:
mode of 'dir.0775' changed from 0755 (rwxr-xr-x) to 0775 (rwxrwxr-x)
drwxrwxr-x 1 root root 0 May 12 06:14 dir.0775

Assumption: cp -Rv respects 0022 umask whether or not the user is root
'dir.0775' -> 'dir.0755/'
Test assumption:
drwxrwxr-x 1 root root 0 May 12 06:14 dir.0755
root@boulder:~/x86_64

hphilm · 2026-05-30T13:42:36Z

Needs a rebase on 0.9.0 for further testing ...

hphilm · 2026-08-16T17:01:17Z

Needs a rebase on 0.10.0 for further testing ...

hphilm · 2026-09-17T07:55:53Z

Most likely superseded by #14576 and/or #14560

my4ng mentioned this pull request Feb 13, 2026

fix(cp): add temporary write/exec perm to built dirs #10859

Closed

ChrisDryden reviewed Feb 13, 2026

View reviewed changes

Comment thread tests/by-util/test_cp.rs Outdated

my4ng force-pushed the cp_mode branch from d65a05e to f23a467 Compare February 21, 2026 13:55

my4ng force-pushed the cp_mode branch from f23a467 to 1d589b4 Compare March 17, 2026 02:24

my4ng force-pushed the cp_mode branch from 305d0c5 to 63d08f3 Compare March 17, 2026 22:50

hphilm mentioned this pull request Mar 24, 2026

uutils-coreutils: Update to v0.8.0 aerynOS/recipes#1532

Closed

2 tasks

my4ng added 6 commits April 16, 2026 16:34

fix(cp): permissions mode and umask aware

8237a28

fix(cp): copy attributes for parents dirs

3539e32

refactor(cp): simplify mode consts

1f6acdb

refactor(cp): remove unnecessary allocations

124d55a

fix(cp): setuid/setgid/sticky bits

c022fbc

test: condense mode tests with rstest

8d082ea

Fix CI-failing checks.

my4ng force-pushed the cp_mode branch from 63d08f3 to 8d082ea Compare April 16, 2026 06:44

Uh oh!

Uh oh!

Conversation

my4ng commented Feb 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Feb 13, 2026

Uh oh!

ChrisDryden commented Feb 13, 2026

Uh oh!

my4ng commented Feb 13, 2026

Uh oh!

github-actions Bot commented Feb 13, 2026

Uh oh!

Uh oh!

github-actions Bot commented Feb 14, 2026

Uh oh!

github-actions Bot commented Feb 21, 2026

Uh oh!

codspeed Bot commented Feb 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging this PR will improve performance by 50.18%

Performance Changes

Footnotes

Uh oh!

my4ng commented Feb 21, 2026

Uh oh!

github-actions Bot commented Mar 17, 2026

Uh oh!

my4ng commented Mar 17, 2026

Uh oh!

github-actions Bot commented Mar 17, 2026

Uh oh!

hphilm commented Apr 14, 2026

Uh oh!

ermo commented Apr 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

hphilm commented Apr 16, 2026

Uh oh!

my4ng commented Apr 16, 2026

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

hphilm commented Apr 16, 2026

Uh oh!

my4ng commented Apr 23, 2026

Uh oh!

hphilm commented May 7, 2026

Uh oh!

hphilm commented May 12, 2026

Uh oh!

hphilm commented May 30, 2026

Uh oh!

hphilm commented Aug 16, 2026

Uh oh!

hphilm commented Sep 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

my4ng commented Feb 13, 2026 •

edited

Loading

codspeed Bot commented Feb 21, 2026 •

edited

Loading

ermo commented Apr 15, 2026 •

edited

Loading