Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] dyndbg: ddebug_find_delimiter() tracks quotes across comments and newlines
--
commit 0c538c4aea6cd34ef1fce80f637f8815b5c63761
Author: Jim Cromie [off-list ref]
dyndbg: split multi-query strings with @
This commit adds '@' as a multi-command delimiter alongside ';' and '\n' in
dynamic debug queries, using a quote-aware helper ddebug_find_delimiter() to
avoid splitting on delimiters inside quoted strings. It also updates the
dynamic debug documentation and selftests accordingly.
quoted hunk ↗ jump to hunk
diff --git a/lib/dynamic_debug.c b/lib/dynamic_debug.c
index 5d99c3aef5a16..11f7350f35907 100644
--- a/lib/dynamic_debug.c
+++ b/lib/dynamic_debug.c
@@ -622,6 +622,30 @@ static int ddebug_exec_query(char *query_string, const char *modname)
return nfound;
}
+/*
+ * Split multi-query string on delimiters (@, ;, \n), ignoring delimiters
+ * that appear inside single or double quoted strings.
+ */
+static char *ddebug_find_delimiter(char *str, const char *delims)
+{
+ bool in_quote = false;
+ char quote_char = 0;
+
+ for (; *str; str++) {
+ if (*str == '"' || *str == '\'') {
+ if (!in_quote) {
+ in_quote = true;
+ quote_char = *str;
+ } else if (*str == quote_char) {
+ in_quote = false;
+ }
+ } else if (!in_quote && strchr(delims, *str)) {
+ return str;
+ }
+ }
+ return NULL;
+}
+
/* handle multiple queries in query string, continue on error, return
last error or number of matching callsites. Module name is either
in the modname arg (for boot args) or perhaps in query string.@@ -632,7 +656,7 @@ static int ddebug_exec_queries(char *query, const char *modname)
int i, errs = 0, exitcode = 0, rc, nfound = 0;
for (i = 0; query; query = split) {
- split = strpbrk(query, ";\n");
+ split = ddebug_find_delimiter(query, "@;\n");
if (split)
*split++ = '\0';
[Severity: Medium]
Can ddebug_find_delimiter() drop or merge subsequent lines in a multi-line
query when a '#' comment contains an apostrophe or when a line has an
unclosed quote?
Previously, ddebug_exec_queries() called strpbrk(query, ";\n"), which
unconditionally split the input buffer on '\n' before comment handling in
ddebug_exec_queries() and ddebug_tokenize():
lib/dynamic_debug.c:ddebug_exec_queries() {
...
query = skip_spaces_and_commas(query);
if (!query || !*query || *query == '#')
continue;
...
}
lib/dynamic_debug.c:ddebug_tokenize() {
...
while (*buf) {
...
if (*buf == '#')
break; /* token starts comment, skip rest of line */
...
}
With ddebug_find_delimiter(query, "@;\n"), quote state is tracked across '#'
comments and across '\n'. If a multi-line write to
/proc/dynamic_debug/control contains an apostrophe inside a comment, such as:
# don't enable foo
func bar +p
or an inline comment:
func foo +p # module's debug
func bar +p
ddebug_find_delimiter() sets in_quote = true on the '\'' and ignores the '\n'
at the end of the comment line. As a result, "func bar +p" is merged into the
commented line and silently discarded when ddebug_exec_queries() or
ddebug_tokenize() skips the comment.
Similarly, if a line has an unclosed quote, ddebug_find_delimiter() no longer
splits at '\n'. On a multi-line buffer, subsequent lines are concatenated
into the broken line instead of failing that line and continuing. On a
single-line write via echo, the trailing '\n' is left attached to the query
string and emitted in the v2pr_info() and pr_err("unclosed quote: %s\n", buf)
messages, which also appears to be why the golden hashes for the unclosed
quote tests (FT_grammar_errs.1, FT_grammar_errs.21, and FT_grammar_errs.41)
changed in GOLDEN_RECORDS in
tools/testing/selftests/dynamic_debug/dyndbg_selftest.sh.
Should '\n' always terminate a query (and reset quote state, or be split
before scanning for ';' and '@'), and should '#' comments outside quotes skip
to the end of the line?
--
Sashiko AI review · https://sashiko.dev/#/patchset/20261027-dd-cmap-part2-clean-v13-0-9267923be612@gmail.com?part=30