Remove the break statement that prevented the file from being closed.

Signed-off-by: Gabriel Adrian Samfira <gsamfira@cloudbasesolutions.com>

The EnableProcessPrivileges function will enable a certain set of
privileges for the entire process. I am unsure if enabling/disabling
privileges is tracked with some sort of reference counting, but if it's
not, when we release those privileges, we disable them for the entire
process. This means that if some other go routine requires them, we may
end up with undesirable results.

The RunWithPrivileges locks the thread in which the function runs, and
enables those privileges only in that thread.

Signed-off-by: Gabriel Adrian Samfira <gsamfira@cloudbasesolutions.com>

Use RunWithPrivileges

Fix leaking file handle

Before:

```
go test -bench=. .
goos: linux
goarch: amd64
pkg: github.com/tonistiigi/fsutil
cpu: Intel(R) Core(TM) i7-10610U CPU @ 1.80GHz
BenchmarkWalker/[1]-target-8         	   27994	     41254 ns/op
BenchmarkWalker/[1]-**/target-8      	   29540	     40402 ns/op
BenchmarkWalker/[2]-*/target-8       	    1017	   1104834 ns/op
BenchmarkWalker/[2]-**/target-8      	     920	   1124731 ns/op
BenchmarkWalker/[3]-*/*/target-8     	      34	  40524094 ns/op
BenchmarkWalker/[3]-**/target-8      	      32	  40432908 ns/op
BenchmarkWalker/[4]-*/*/*/target-8   	      31	  44836714 ns/op
BenchmarkWalker/[4]-**/target-8      	      26	  44611379 ns/op
BenchmarkWalker/[5]-*/*/*/*/target-8 	      85	  14179975 ns/op
BenchmarkWalker/[5]-**/target-8      	      78	  13620031 ns/op
BenchmarkWalker/[6]-*/*/*/*/*/target-8         	      44	  23380013 ns/op
BenchmarkWalker/[6]-**/target-8                	      51	  22435264 ns/op
BenchmarkWalker/[6]-**-!*/*/**-8               	    2101	    580936 ns/op
```

After:

```
go test -bench=. .
goos: linux
goarch: amd64
pkg: github.com/tonistiigi/fsutil
cpu: Intel(R) Core(TM) i7-10610U CPU @ 1.80GHz
BenchmarkWalker/[1]-target-8         	   40788	     26219 ns/op
BenchmarkWalker/[1]-**/target-8      	   41642	     26998 ns/op
BenchmarkWalker/[2]-*/target-8       	    1725	    660271 ns/op
BenchmarkWalker/[2]-**/target-8      	    1806	    645525 ns/op
BenchmarkWalker/[3]-*/*/target-8     	      64	  17609101 ns/op
BenchmarkWalker/[3]-**/target-8      	      66	  17563334 ns/op
BenchmarkWalker/[4]-*/*/*/target-8   	      40	  26241578 ns/op
BenchmarkWalker/[4]-**/target-8      	      43	  24575370 ns/op
BenchmarkWalker/[5]-*/*/*/*/target-8 	     121	   9564283 ns/op
BenchmarkWalker/[5]-**/target-8      	     126	   9412483 ns/op
BenchmarkWalker/[6]-*/*/*/*/*/target-8         	      63	  18703020 ns/op
BenchmarkWalker/[6]-**/target-8                	      73	  17287584 ns/op
BenchmarkWalker/[6]-**-!*/*/**-8               	    3792	    273148 ns/op
```

Signed-off-by: Nick Santos <nick.santos@docker.com>

If using the diskwriter hanlders in parallel we might see a common
directory being created in parallel, only one will win.  This allows us
to gracefully retry/re-evaluate when we get a syscall.EEXIST error on
the mkdir instead of returning an error.

handle mkdir race for diskwriter

walk: avoid stat()'ing files unnecessarily

Signed-off-by: CrazyMax <crazy-max@users.noreply.github.com>

Signed-off-by: CrazyMax <crazy-max@users.noreply.github.com>

Signed-off-by: CrazyMax <crazy-max@users.noreply.github.com>

Signed-off-by: CrazyMax <crazy-max@users.noreply.github.com>

ci: fix freebsd workflow

Signed-off-by: Tonis Tiigi <tonistiigi@gmail.com>

update to Go 1.20