Skip to contentSkip to conversation

scripts/checkpatch: Match paths beginning with hw/

PATCH New TANG Tiancheng 1 participant 1 message Aug 4 Checkpatch

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 scripts/checkpatch: Match paths beginning with hw/

Patch mode strips the leading a/ and b/ components from file paths, leaving repository-relative paths such as hw/riscv/foo.c. The checks for qemu_bh_new() and aio_bh_new() currently require /hw/ in $realfile, so they miss paths beginning with hw/ in both normal Git patches and --file mode.

Match either the start of $realfile or a slash before hw/, making patch and file modes consistent while preserving nested /hw/ matches.

Fixes ef56ffbdd6b0 ("checkpatch: add qemu_bh_new/aio_bh_new checks")

---
The checks for qemu_bh_new() and aio_bh_new() introduced by ef56ffbdd6b0 are intended to cover files under hw/. In patch mode, checkpatch removes Git's leading a/ and b/ path components before setting $realfile. A file under the repository's hw/ directory is therefore represented as hw/foo.c. Since the checks require /hw/, they do not recognize that path. The same false negative occurs in --file mode when given a repository-relative path such as hw/misc/i2c-echo.c.

The problem can be reproduced from an unmodified upstream QEMU checkout at the series base:

  $ git show b14037f37f9 -- hw/misc/i2c-echo.c |
      scripts/checkpatch.pl --patch \
          --test-only=aio_bh_new_guarded --no-summary -
  Your patch has no obvious style problems and is ready for submission.

Commit b14037f37f9 adds a qemu_bh_new() call under hw/, so that result is a false negative. Repository-relative file mode reproduces the issue:

  $ scripts/checkpatch.pl --file hw/misc/i2c-echo.c \
      --test-only=aio_bh_new_guarded --no-summary
  hw/misc/i2c-echo.c has no obvious style problems and is ready for
  submission.

Before this patch, both invocations succeed without reporting the call. After this patch, both report:

ERROR: use aio_bh_new_guarded() instead of qemu_bh_new*() to avoid reentrancy problems

and return failure as expected. The analogous aio_bh_new() check was verified with hw/i386/kvm/xen_evtchn.c.

---
 scripts/checkpatch.pl | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
scripts/checkpatch.pl +4 -2
index 03f35e75012c8c7a00b67f036e4befb2ea162d4a..9ae7693a4f388597dc11a58a85736d511fbcc62c 100755
@@ -3223,11 +3223,13 @@ sub process {
 			ERROR("use sigaction to establish signal handlers; signal is not portable\n" . $herecurr);
 		}
 # recommend aio_bh_new_guarded instead of legacy qemu_bh_new / qemu_bh_new_guarded
-        if ($realfile =~ /.*\/hw\/.*/ && $line =~ /\bqemu_bh_new(_guarded)?\s*\(/) {
+		if ($realfile =~ m{(?:^|/)hw/} &&
+		    $line =~ /\bqemu_bh_new(_guarded)?\s*\(/) {
 			ERROR("use aio_bh_new_guarded() instead of qemu_bh_new*() to avoid reentrancy problems\n" . $herecurr);
 		}
 # recommend aio_bh_new_guarded instead of aio_bh_new
-        if ($realfile =~ /.*\/hw\/.*/ && $line =~ /\baio_bh_new\s*\(/) {
+		if ($realfile =~ m{(?:^|/)hw/} &&
+		    $line =~ /\baio_bh_new\s*\(/) {
 			ERROR("use aio_bh_new_guarded() instead of aio_bh_new() to avoid reentrancy problems\n" . $herecurr);
 		}
 # check for DEVICE_NATIVE_ENDIAN, use explicit endianness instead

---
base-commit: b428fe036233cbd15d37e3c027ab6ca4d3661a80
change-id: 20260804-b4-checkpatch-hw-path-cd477b5357b8

Best regards,

Signature · 1 line
TANG Tiancheng <[email protected]>