mv: create cross-device directory entries relative to the destination descriptor by abendrothj · Pull Request #14971 · uutils/coreutils

abendrothj · 2026-09-29T23:50:48Z

When a directory move falls back to copying across filesystems, each destination entry was created by joining a path, and regular files went through fs::copy. A symlink that appeared in the destination during the copy was followed, so content went wherever it pointed.

Each destination directory is now held open (search-only where possible), and files, subdirectories, symlinks, special files and hard links are created relative to it without following symlinks. An entry that shows up at a name is replaced through replace_entry_at. Regular files share copy_file_data with the single-file fallback, which sets ownership and mode on the open descriptor, using libc fchown so fakeroot still works. Source directory names are read before descending, so one descriptor stays open per level.

Still by path: directory ownership, symlink xattrs, the hard link source, and the final directory xattr reopen. Run on macOS and Linux. On Debian GNU/Hurd the mv tests fail only where main does (the new ones are Linux-only), and a cross-device move of a 0640 FIFO, alone or in a directory, keeps its mode as with GNU 9.10. AIX is only cross-checked with clippy.

Based on #15123; until the stack below it merges, the diff here shows those commits too.

Closes #14991

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

codspeed · 2026-09-30T00:19:56Z

Merging this PR will degrade performance by 10.68%

⚡ 7 improved benchmarks
❌ 13 regressed benchmarks
✅ 379 untouched benchmarks
⏩ 54 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	`BASE`	`HEAD`	Efficiency
❌	Simulation	`three_39_bit_primes`	371.3 ms	613.8 ms	-39.5%
❌	Memory	`dd_copy_default`	19.7 KB	28.6 KB	-31.08%
❌	Memory	`dd_copy_4k_blocks`	23.1 KB	31.9 KB	-27.83%
❌	Memory	`dd_copy_partial`	23.1 KB	32 KB	-27.82%
❌	Memory	`dd_copy_with_seek`	23.4 KB	32.3 KB	-27.56%
❌	Memory	`dd_copy_with_skip`	23.4 KB	32 KB	-26.89%
❌	Memory	`dd_copy_8k_blocks`	27.1 KB	35.6 KB	-24.1%
❌	Simulation	`five_38_bit_primes`	1.7 s	2.2 s	-21.62%
❌	Simulation	`true_consecutive_calls`	294.7 ns	350.2 ns	-15.86%
❌	Memory	`dd_copy_64k_blocks`	83.1 KB	91.9 KB	-9.67%
❌	Simulation	`dd_copy_partial`	717.6 µs	755.5 µs	-5.01%
❌	Simulation	`hostname_ip_lookup[100000]`	169.1 µs	178 µs	-4.99%
❌	Memory	`dd_copy_separate_blocks`	185.4 KB	194.8 KB	-4.83%
⚡	Simulation	`split_lines`	11.1 ms	8.5 ms	+29.92%
⚡	Simulation	`split_numeric_suffix`	11.3 ms	8.7 ms	+29.23%
⚡	Simulation	`tsort_complex_dag[50000]`	97.2 ms	87.6 ms	+10.86%
⚡	Simulation	`tsort_tree_dag[(10, 3)]`	39.4 ms	36.5 ms	+7.93%
⚡	Simulation	`tsort_wide_dag[100000]`	167 ms	157.6 ms	+5.98%
⚡	Simulation	`tsort_linear_chain[1000000]`	2 s	1.9 s	+4.99%
⚡	Simulation	`thirteen_39_bit_primes`	9 s	8.6 s	+4.52%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

_{Comparing abendrothj:mv-cross-device-dir-dirfd (b3b9dde) with main (8cf2e4f)}

54 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. ↩

github-actions · 2026-09-30T00:44:15Z

GNU testsuite comparison:

Congrats! The gnu test tests/mv/mv-special-1 is no longer failing!

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

abendrothj · 2026-10-01T00:53:50Z

Opened #14991.

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

sylvestre · 2026-10-05T21:25:46Z

this is quite a large patch, can we make it smaller?

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

abendrothj · 2026-10-06T07:29:44Z

Smaller now: it reuses DirFd, replace_entry_at, link_at and copy_special_file from the PRs below it instead of its own per-entry helpers. Its own diff went from +588/-66 to 5 files, +490/-79, about 200 of them tests. Part of that is a late fix in 7e5ee68: ownership is set with libc fchown on the open descriptor, so fakeroot can intercept it, with a regression test whose prerequisites are in DEVELOPMENT.md. It sits on #15123, so the diff shown here includes the stack below it until that lands.

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

abendrothj · 2026-10-07T00:55:11Z

drafting this until #15120, #15122 and #13334 land; on its own it's +490/-79, about 200 of them tests.

codspeed · 2026-10-07T03:41:19Z

Unable to generate the flame graphs

The performance report has correctly been generated, but there was an internal error while generating the flame graphs for this run. We're working on fixing the issue. Feel free to contact us on Discord or at support@codspeed.io if the issue persists.

Open a directory readable, and on EACCES retry with O_PATH or O_SEARCH, which anchor *at calls without read access. The targets with either flag are named by the has_o_path and has_o_search cfg aliases. If the retry fails, its own error is returned.

Move the temporary name and renameat logic of replace_link into replace_entry_at, which creates the entry through a callback relative to an open directory, so other callers can replace entries the same way. The parent is now opened with DirFd::open_anchor, so ln -sf replacing a link in a directory with write and search but no read permission works, as it does with GNU, instead of failing with EACCES.

Create fifos, sockets and device nodes relative to an open directory. nix and rustix do not provide mknodat on Apple targets, so this calls libc directly.

A cross-device move of a socket or device node went through the regular file copy, which removed the destination and then failed to open the source. A fifo also removed the destination before creating the new one, and lost its mode. Fifos, sockets and device nodes are now created with their mode and ownership in a private directory next to the destination and renamed over it, so the destination is kept if the node cannot be created, and the ownership and mode cannot land on another file linked over the destination name meanwhile. The private directory is made 0700 by name right after it is created, before it is opened, without changing the umask; the check after the open rejects a directory moved there in between. As for regular files, setuid and setgid are dropped when the ownership cannot be kept. Fixes uutils#13145

Sockets and device nodes inside a directory moved across devices went through the regular file copy, which cannot open them, so the move failed. Fifos were recreated with mode 0666 minus the umask. Recreate all of them with copy_special_file, which keeps the mode.

A caller creating a file can then keep writing through the descriptor it opened. Also make link_at public, for creating links relative to an open directory, and add DirFd::open_subdir_anchor, which opens a subdirectory without following a symlink and, where the platform allows, without needing read permission, for creating entries in it.

When a directory move falls back to copying across filesystems, each destination entry was created by joining a path, and regular files went through fs::copy. A symlink that appeared in the destination while the copy was running was followed, so the content went wherever it pointed. Hold each destination directory open, search-only where possible, and create files, subdirectories, symlinks, special files and hard links relative to it without following symlinks. An entry that appears at a name is replaced through replace_entry_at. Regular files share copy_file_data with the single file fallback, which sets ownership and mode through the descriptor. The names of each source directory are read before descending, so only one descriptor stays open per level. Closes uutils#14991

Copilot AI balanced review requested due to automatic review settings September 29, 2026 23:50

Copilot AI reviewed Sep 29, 2026

abendrothj force-pushed the mv-cross-device-dir-dirfd branch from d0cf9fe to f3fc847 Compare September 29, 2026 23:54

Copilot AI balanced review requested due to automatic review settings September 29, 2026 23:54

Copilot AI reviewed Sep 29, 2026

abendrothj force-pushed the mv-cross-device-dir-dirfd branch from f3fc847 to e3fda2a Compare September 29, 2026 23:55

abendrothj force-pushed the mv-cross-device-dir-dirfd branch from e3fda2a to 8d392d1 Compare September 30, 2026 02:35

Copilot AI balanced review requested due to automatic review settings September 30, 2026 02:35

Copilot AI reviewed Sep 30, 2026

Copilot AI balanced review requested due to automatic review settings October 1, 2026 01:14

abendrothj force-pushed the mv-cross-device-dir-dirfd branch from 8d392d1 to 5e3cb57 Compare October 1, 2026 01:14

Copilot AI reviewed Oct 1, 2026

Cassian433 mentioned this pull request Oct 4, 2026

mv: keep copying xattrs after one fails in a cross-device move #15049

Merged

abendrothj force-pushed the mv-cross-device-dir-dirfd branch from 5e3cb57 to 42731bd Compare October 5, 2026 06:13

Copilot AI balanced review requested due to automatic review settings October 5, 2026 06:13

Copilot AI reviewed Oct 5, 2026

This was referenced Oct 5, 2026

mv: no warning when the destination file system doesn't support xattrs (GNU cp-mv-enotsup-xattr, 9.12) #15079

Open

mv: warn when the destination doesn't support xattrs #15101

Open

Copilot AI balanced review requested due to automatic review settings October 6, 2026 07:15

abendrothj force-pushed the mv-cross-device-dir-dirfd branch from 42731bd to 7e5ee68 Compare October 6, 2026 07:15

Copilot AI reviewed Oct 6, 2026

Copilot AI balanced review requested due to automatic review settings October 6, 2026 07:58

Copilot AI reviewed Oct 6, 2026

abendrothj force-pushed the mv-cross-device-dir-dirfd branch from b125862 to b3b9dde Compare October 7, 2026 00:54

Copilot AI balanced review requested due to automatic review settings October 7, 2026 00:54

Copilot AI reviewed Oct 7, 2026

abendrothj marked this pull request as draft October 7, 2026 00:55

abendrothj added 8 commits October 7, 2026 01:05

uucore: format the fs feature list for taplo

5d10e9b

uucore: add DirFd::mknod_at

0fb05fd

Create fifos, sockets and device nodes relative to an open directory. nix and rustix do not provide mknodat on Apple targets, so this calls libc directly.

abendrothj force-pushed the mv-cross-device-dir-dirfd branch from b3b9dde to 52022fc Compare October 7, 2026 08:21

Conversation

abendrothj commented Sep 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

codspeed Bot commented Sep 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging this PR will degrade performance by 10.68%

Performance Changes

Footnotes

Uh oh!

github-actions Bot commented Sep 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

abendrothj commented Oct 1, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

sylvestre commented Oct 5, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

abendrothj commented Oct 6, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

abendrothj commented Oct 7, 2026

Uh oh!

codspeed Bot commented Oct 7, 2026

Unable to generate the flame graphs

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

abendrothj commented Sep 29, 2026 •

edited

Loading

codspeed Bot commented Sep 30, 2026 •

edited

Loading

github-actions Bot commented Sep 30, 2026 •

edited

Loading