ls: Reduce xattr calls

[kimono-koans]

Master calls listxattr more times than necessary:

% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ------------------
 17.83    0.000783         783         1           execve
 15.87    0.000697          15        44           getdents64
 11.82    0.000519           6        84           listxattr
 10.00    0.000439           8        53         6 statx
  8.72    0.000383          12        31         1 openat
  6.67    0.000293          22        13           mmap
  5.05    0.000222           6        34           close
  4.94    0.000217          19        11           read
  4.19    0.000184           6        30           fstat
  1.73    0.000076          15         5           mprotect
  1.53    0.000067          33         2           munmap
  1.46    0.000064          16         4         4 connect
  1.30    0.000057           9         6           brk
  1.09    0.000048          12         4           socket
  1.07    0.000047          47         1           write
  0.89    0.000039           7         5           ioctl
  0.82    0.000036           9         4           newfstatat
  0.77    0.000034           3         9           rt_sigaction
  0.59    0.000026          26         1           set_robust_list
  0.48    0.000021          10         2           pread64
  0.46    0.000020           6         3           fcntl
  0.43    0.000019          19         1         1 access
  0.32    0.000014           7         2           prlimit64
  0.30    0.000013          13         1           poll
  0.30    0.000013           1         9         6 readlink
  0.27    0.000012           4         3           sigaltstack
  0.23    0.000010           3         3           lseek
  0.23    0.000010          10         1           getrandom
  0.20    0.000009           9         1           sched_getaffinity
  0.16    0.000007           7         1           arch_prctl
  0.16    0.000007           7         1           set_tid_address
  0.14    0.000006           6         1           rseq
  0.00    0.000000           0         1           gettid
------ ----------- ----------- --------- --------- ------------------
100.00    0.004392          11       372        18 total

This PR reduces those calls:

% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ------------------
 19.28    0.000822         822         1           execve
 11.75    0.000501          11        42           llistxattr
  8.66    0.000369           8        44           getdents64
  8.02    0.000342           6        53         6 statx
  7.62    0.000325          10        31         1 openat
  7.55    0.000322          24        13           mmap
  5.98    0.000255          51         5           mprotect
  4.60    0.000196          17        11           read
  4.41    0.000188           5        34           close
  4.03    0.000172           5        30           fstat
  2.63    0.000112          12         9         6 readlink
  2.42    0.000103          25         4         4 connect
  2.18    0.000093          23         4           socket
  1.81    0.000077          12         6           brk
  1.55    0.000066          33         2           munmap
  1.31    0.000056           6         9           rt_sigaction
  0.89    0.000038          38         1           write
  0.84    0.000036           9         4           newfstatat
  0.61    0.000026          13         2           pread64
  0.47    0.000020           6         3           fcntl
  0.42    0.000018          18         1         1 access
  0.38    0.000016           5         3           sigaltstack
  0.35    0.000015           7         2           prlimit64
  0.30    0.000013          13         1           poll
  0.30    0.000013           4         3           lseek
  0.28    0.000012           2         5           ioctl
  0.23    0.000010          10         1           sched_getaffinity
  0.21    0.000009           9         1           arch_prctl
  0.21    0.000009           9         1           getrandom
  0.19    0.000008           8         1           set_tid_address
  0.19    0.000008           8         1           rseq
  0.16    0.000007           7         1           set_robust_list
  0.14    0.000006           6         1           gettid
------ ----------- ----------- --------- --------- ------------------
100.00    0.004263          12       330        18 total

[codspeed-hq]

Merging this PR will improve performance by 4.21%

⚡ 3 improved benchmarks
✅ 294 untouched benchmarks
⏩ 48 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	ls_recursive_long_all_mixed_tree	2.7 ms	2.6 ms	+3.58%
⚡	Simulation	ls_recursive_long_all_balanced_tree[(6, 4, 15)]	133.5 ms	128.5 ms	+3.86%
⚡	Simulation	ls_recursive_long_all_deep_tree[(100, 4)]	2.9 ms	2.8 ms	+4.21%

---

_{Comparing kimono-koans:excessive_xattr_calls (615bf4b) with main (d830e84)}

Footnotes

1. 48 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. ↩

[sylvestre]

Interesting that it regressed the performances

[kimono-koans]

Interesting that it regressed the performances

Better now that we only request xattrs once!

% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ------------------
 35.84    0.002141          48        44           getdents64
 11.33    0.000677          21        31         1 openat
 11.23    0.000671         671         1           execve
  8.47    0.000506           9        54         6 statx
  5.51    0.000329          25        13           mmap
  5.11    0.000305           8        34           close
  4.52    0.000270           9        30           fstat
  3.75    0.000224          20        11           read
  2.21    0.000132          33         4         4 connect
  1.76    0.000105          21         5           mprotect
  1.71    0.000102          17         6           brk
  1.67    0.000100          11         9         6 readlink
  1.61    0.000096          24         4           socket
  0.79    0.000047          11         4           newfstatat
  0.64    0.000038          19         2           munmap
  0.64    0.000038           4         9           rt_sigaction
  0.54    0.000032           6         5           ioctl
  0.42    0.000025          25         1           set_robust_list
  0.39    0.000023          23         1           llistxattr
  0.25    0.000015           7         2           pread64
  0.23    0.000014          14         1         1 access
  0.22    0.000013           4         3           fcntl
  0.18    0.000011          11         1           poll
  0.18    0.000011           3         3           lseek
  0.17    0.000010           5         2           prlimit64
  0.15    0.000009           3         3           sigaltstack
  0.12    0.000007           7         1           sched_getaffinity
  0.12    0.000007           7         1           getrandom
  0.07    0.000004           4         1           arch_prctl
  0.07    0.000004           4         1           gettid
  0.07    0.000004           4         1           set_tid_address
  0.07    0.000004           4         1           rseq
  0.00    0.000000           0         1           write
------ ----------- ----------- --------- --------- ------------------
100.00    0.005974          20       290        18 total

[Alonely0]

@kimono-koans hey kimono, I wanna apologize for submitting the refactor PR, I missed yours and now we have this conflict mess. Let me know if you're still interested in pursuing this PR or want me to fix the merge conflicts, I'll gladly do it :)

Elsewhere, I'd be interested in picking up your optimizations here and cleanly applying them to the new HEAD (with proper attribution, of course), if that's good by you.

[sylvestre]

Looks good to me
Please preserve the author in the commit

[kimono-koans]

@kimono-koans hey kimono, I wanna apologize for submitting the refactor PR, I missed yours and now we have this conflict mess. Let me know if you're still interested in pursuing this PR or want me to fix the merge conflicts, I'll gladly do it :)

No problem!

Elsewhere, I'd be interested in picking up your optimizations here and cleanly applying them to the new HEAD (with proper attribution, of course), if that's good by you.

@Alonely0 Yes, please feel free to move ahead with this PR without me. I feel like I would need a mouse and monitor and time to refactor, and I don't have any of those right now!

@sylvestre Pleased to coauthor or whatever. Whatever moves this/us forward.

See also: #11423