perf DSO hardening series
Arnaldo Carvalho de Melo
maintainerv4 cover perf DSO hardening series
Hi,
Please consider merging,
Five hardening fixes for the perf DSO data handling code path, all found by the sashiko-bot AI reviewer:
- __open_dso() can return a file descriptor of 0 (stdin) when dso__get_filename() fails without setting errno, so fd = -errno = 0 looked like a successfully opened DSO;
- dso__decompress_kmodule_path() calls close(fd) with fd = -1 when decompression fails or the DSO is not compressed, clobbering errno with EBADF on the error path;
- file_read() and file_size() read the stale global errno after try_to_open_dso() fails, when by then errno has been through mutex_lock(), nsinfo__mountns_enter() and several open() attempts;
- dso_cache__memcpy() underflows when cache->size is smaller than the
RB tree offset (short pread), computing a huge memcpy length;
- dso__read_symbol() asserts on an input-dependent condition instead of doing a runtime check, keeping the BPF metadata safe from crafted input.
Best regards,
- Arnaldo
What changed from v3 (41ed39bf2dac1d80):
PATCH 3/5:
- Added assert(ret < 0) after ret = dso__data(dso)->fd in both file_read() and file_size() to verify the comment that fd is always negative, never 0 [Ian Rogers review nit].
- All 5 patches carry Reviewed-by: Ian Rogers.
What changed from v2 (2f944ea022b6429f):
PATCH 3/5:
- file_size() now also uses the stored fd error instead of the stale global errno on open failure; the commit was retitled to cover both file_read() and file_size() [sashiko-bot review of PATCH 3/5].
- Comment corrected: on failure dso__data(dso)->fd is always negative — -errno from __open_dso() or -1 from do_open() — never 0.
PATCH 4/5:
- Clarified that returning 0 for an offset past a short-read chunk is EOF semantics — for a regular file a short pread only happens at end-of-file, so 0 is what a direct pread() at that offset would return — not a cache-miss that triggers a re-read; comment and commit message updated [sashiko-bot review of PATCH 4/5].
- The lockless dso cache RB tree lookup vs. concurrent insert question raised in the same review is pre-existing; it is recorded in tools/perf/TODO.hardening (item 175) for follow-up work, with no change to this series.
What changed from v1 ([email protected]):
PATCH 1/5:
- __open_dso() now sets errno = ENOENT directly when dso__get_filename() fails with errno == 0, instead of only computing fd = -ENOENT. Callers that check errno after a negative fd (e.g. file_read() returning -errno) no longer get 0/EOF for an open failure [sashiko-bot review of PATCH 1/5].
- Rebased onto the current perf-tools-next head (bf10e6ee2ac3034c).
This series was developed with assistance from Claude (claude-opus-4.6) and Opencode (deepseek-v4-flash-free) for code analysis, patch generation, and commit message composition. All changes were validated by the maintainer.
Arnaldo Carvalho de Melo (5):
perf dso: Guard against errno==0 when dso__get_filename() returns NULL perf dso: Guard close() against invalid fd in
dso__decompress_kmodule_path()
perf dso: Use stored fd error instead of stale errno in file_read()
and file_size()
perf dso: Guard against cache underflow on short reads in
dso_cache__memcpy()
perf dso: Replace assert with runtime check in dso__read_symbol()
tools/perf/util/dso.c | 41 ++++++++++++++++++++++++++++++++++------- 1 file changed, 34 insertions(+), 7 deletions(-)
base-commit: bf10e6ee2ac3034c9068e03eed418fd16961984e
v3-head: 41ed39bf2dac1d80
v2-head: 2f944ea022b6429f
v1-head: f6cb9e46c7b8949b
Arnaldo Carvalho de Melo
maintainerPATCH v4 1/5 perf dso: Guard against errno==0 when dso__get_filename() returns NULL
From: Arnaldo Carvalho de Melo <[email protected]>
__open_dso() computes fd = -errno when dso__get_filename() returns NULL. Some failure paths in dso__get_filename() (e.g. binary type mismatch) return NULL without making a syscall, leaving errno at 0 from a prior successful call. fd = -0 = 0, which is stdin — subsequent code treats it as a valid file descriptor.
Fall back to ENOENT when errno is 0, ensuring fd is always negative on failure.
--- tools/perf/util/dso.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
tools/perf/util/dso.c +5 -2
@@ -640,10 +640,13 @@ static int __open_dso(struct dso *dso, struct machine *machine)mutex_lock (dso__lock (dso)); name =dso__get_filename (dso, machine ? machine->root_dir :"" , &decomp); -if (name) +if (name) { fd =do_open (name); -else + }else { +if (errno ==0 ) + errno =ENOENT ; fd = -errno; + }if (decomp)unlink (name);
Signature · 1 line
2.55.0
[email protected]
AI reviewRe: [PATCH 1/5] perf dso: Guard against errno==0 when dso__get_filename() returns NULL
Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
New issues:
- [High] Stale errno usage in dso__get_filename() and __open_dso() creates brittle logic, incorrect error propagation, and loop pollution.
Pre-existing issues:
- [High] file_read() attempts to extract the failure reason using -errno after several cleanup and namespace functions have clobbered the errno value.
- [High] Strict AB-BA deadlock between _dso__data_open_lock and dso__lock when REFCNT_CHECKING is enabled and the EMFILE limit is hit. --
--- Patch [1]: [PATCH 1/5] perf dso: Guard against errno==0 when dso__get_filename() returns NULL --- commit b3606e9e10d4ce3d99991eea12437267b852da75
Author: Arnaldo Carvalho de Melo <[email protected]>
perf dso: Guard against errno==0 when dso__get_filename() returns NULL
This commit fixes an issue where __open_dso() could incorrectly return 0 (stdin) when dso__get_filename() returns NULL without setting errno. It forces errno to ENOENT if it is 0 upon failure, ensuring a negative file descriptor is returned.
perf/util/dso.c static int __open_dso(struct dso *dso, struct machine *machine) · line 640
Quoted 4 lines
perf/util/dso.c +0
-0@@ -640,10 +640,13 @@ static int __open_dso(struct dso *dso, struct machine *machine)mutex_lock (dso__lock (dso));
[Severity: High]
This is a pre-existing issue, but can this lock ordering cause an AB-BA deadlock when REFCNT_CHECKING is enabled?
If do_open() hits the EMFILE limit, close_first_dso() calls dso__list_del(). Under REFCNT_CHECKING, dso__list_del() drops and reacquires dso__data_open_lock() to satisfy constraints:
tools/perf/util/dso.c:dso__list_del() {
#ifdef REFCNT_CHECKING
mutex_unlock(dso__data_open_lock());
dso__put(dso__data(dso)->dso);
mutex_lock(dso__data_open_lock());#endif
}
Since dso__lock(dso) is held here in __open_dso(), if another thread acquires dso__data_open_lock() and then attempts __open_dso() on the same dso, will they deadlock waiting on each other?
perf/util/dso.c static int __open_dso(struct dso *dso, struct machine *machine) · line 646
Quoted 9 lines
perf/util/dso.c +4
-2name =dso__get_filename (dso, machine ? machine->root_dir :"" , &decomp); -if (name) +if (name) { fd =do_open (name); -else + }else { +if (errno ==0 ) + errno =ENOENT ;
[Severity: High]
Does setting errno to ENOENT here pollute the thread-local state for subsequent loop iterations?
If a missing DSO file triggers a fallback loop like try_to_open_dso(), and the next iteration evaluates a directory instead of an ELF file, is_regular_file() will return false.
Since stat() succeeds on a directory, it won't change errno. The stale ENOENT from here means dso__get_filename() evaluates if (errno != ENOENT || ...) as false.
Will this cause the code to incorrectly bypass the error path, proceed to dso__filename_with_chroot(), and successfully open a directory instead of an ELF file?
perf/util/dso.c static int __open_dso(struct dso *dso, struct machine *machine) · line 648
Quoted 2 lines
perf/util/dso.c +2
-0+ fd = -errno; + }
[Severity: High]
This is a pre-existing issue, but does relying on -errno here and in file_read() lead to incorrect error propagation?
When try_to_open_dso() fails, file_read() attempts to extract the failure reason:
tools/perf/util/dso.c:file_read() {
...
if (dso__data(dso)->fd < 0) {
dso__data(dso)->status = DSO_DATA_STATUS_ERROR;
ret = -errno;
goto out;
}
}However, between the actual failure and this check, cleanup functions like unlink(), mutex_unlock(), and nsinfo__mountns_exit() execute.
Will these calls clobber errno, causing file_read() to return an arbitrary error code or 0, silently failing to report the true failure reason?
Quoted 7 lines
perf/util/dso.c +0
-0if (decomp)unlink (name);mutex_unlock (dso__lock (dso));free (name);return fd; }
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Arnaldo Carvalho de Melo
maintainerPATCH v4 2/5 perf dso: Guard close() against invalid fd in dso__decompress_kmodule_path()
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
dso__decompress_kmodule_path() unconditionally calls close(fd) on the return value of decompress_kmodule(). When decompression fails or the DSO is not compressed, decompress_kmodule() returns -1. close(-1) fails with EBADF and clobbers errno, which callers up the chain (dso__get_filename → __open_dso) depend on for error propagation.
Guard the close() call with fd >= 0 so only valid file descriptors are closed.
--- tools/perf/util/dso.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
tools/perf/util/dso.c +3 -1
@@ -395,7 +395,9 @@ int dso__decompress_kmodule_path(struct dso *dso, const char *name, {int fd =decompress_kmodule (dso, name, pathname, len); -close (fd); +/* decompress_kmodule() returns -1 on failure, don't close(-1) */ +if (fd >=0 ) +close (fd);return fd >=0 ?0 :-1 ; }
Signature · 1 line
2.55.0
[email protected]
AI reviewRe: [PATCH 2/5] perf dso: Guard close() against invalid fd in dso__decompress_kmodule_path()
Sashiko has reviewed this patch and found no issues. It looks great!
Quoted 8 lines
From: Arnaldo Carvalho de Melo <[email protected]>
dso__decompress_kmodule_path()unconditionally calls close(fd) on the return value ofdecompress_kmodule(). When decompression fails or the DSO is not compressed,decompress_kmodule()returns -1. close(-1) fails with EBADF and clobbers errno, which callers up the chain (dso__get_filename→__open_dso) depend on for error propagation.
Quoted 8 lines
Guard the
close()call with fd >= 0 so only valid file descriptors are closed.Fixes 42b3fa670825 ("perf tools: Introduce dso__decompress_kmodule_{fd,path}")Reported-by sashiko-bot <[email protected]>Cc Namhyung Kim <[email protected]>Assisted-by Claude:claude-opus-4.6Signed-off-by Arnaldo Carvalho de Melo <[email protected]>
Sashiko has reviewed this patch and found no issues. It looks great!
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2
Arnaldo Carvalho de Melo
maintainerPATCH v4 3/5 perf dso: Use stored fd error instead of stale errno in file_read() and file_size()
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
file_read() and file_size() use ret = -errno when dso__data(dso)->fd is negative after try_to_open_dso() fails. By this point errno has been through mutex_lock(), nsinfo__mountns_enter(), and multiple open() attempts inside try_to_open_dso() — it no longer reflects the actual open failure. If errno happens to be 0, ret = 0 looks like EOF rather than an error, and file_size() callers like dso__data_size() would then report a zero-sized file instead of failing.
dso__data(dso)->fd is always negative on failure — -errno from __open_dso() when no filename could be built (e.g. -EINVAL, -ENOENT), or -1 when do_open() itself failed — and never 0, so use it directly instead of reading the stale global errno.
--- tools/perf/util/dso.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-)
tools/perf/util/dso.c +6 -2
@@ -1029,7 +1029,9 @@ static ssize_t file_read(struct dso *dso, struct machine *machine,if (dso__data (dso)->fd <0 ) {dso__data (dso)->status =DSO_DATA_STATUS_ERROR ; - ret = -errno; +/* fd is always negative here: -errno or -1, never 0 */ + ret =dso__data (dso)->fd; +assert (ret <0 ); goto out; }
@@ -1151,8 +1153,10 @@ static int file_size(struct dso *dso, struct machine *machine)try_to_open_dso (dso, machine);if (dso__data (dso)->fd <0 ) { - ret = -errno;dso__data (dso)->status =DSO_DATA_STATUS_ERROR ; +/* fd is always negative here: -errno or -1, never 0 */ + ret =dso__data (dso)->fd; +assert (ret <0 ); goto out; }
Signature · 1 line
2.55.0
[email protected]
AI reviewRe: [PATCH 3/5] perf dso: Use stored fd error instead of stale errno in file_read() and file_size()
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Low] Missing explicit inclusion of `<assert.h>` for the newly added `assert()` calls, violating musl…
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] Missing explicit inclusion of <assert.h> for the newly added assert() calls, violating musl libc compatibility guidelines. --
--- Patch [3]: [PATCH 3/5] perf dso: Use stored fd error instead of stale errno in file_read() and file_size() --- commit 9359e4a3d3517acb5c94e232e0f7e3c94b3a7824
Author: Arnaldo Carvalho de Melo <[email protected]>
perf dso: Use stored fd error instead of stale errno in file_read() and file_size()
This commit updates file_read() and file_size() to use the negative fd value as the error code directly instead of relying on errno. This prevents issues where a stale errno could incorrectly appear as an EOF rather than an open failure.
tools/perf/util/dso.c static ssize_t file_read(struct dso *dso, struct machine *machine, · line 1034
Quoted 12 lines
tools/perf/util/dso.c +3
-1@@ -1029,7 +1029,9 @@ static ssize_t file_read(struct dso *dso, struct machine *machine,if (dso__data (dso)->fd <0 ) {dso__data (dso)->status =DSO_DATA_STATUS_ERROR ; - ret = -errno; +/* fd is always negative here: -errno or -1, never 0 */ + ret =dso__data (dso)->fd; +assert (ret <0 );
[Severity: Low]
Does this missing include break musl libc compatibility?
The perf subsystem guidelines require explicit inclusion of POSIX standard headers for any libc macros used. Could you add the <assert.h> header at the top of tools/perf/util/dso.c to ensure this compiles properly under musl?
Quoted 2 lines
tools/perf/util/dso.c +0
-0goto out; }
[ ... ]
tools/perf/util/dso.c static int file_size(struct dso *dso, struct machine *machine) · line 1159
Quoted 9 lines
tools/perf/util/dso.c +3
-1@@ -1151,8 +1153,10 @@ static int file_size(struct dso *dso, struct machine *machine)try_to_open_dso (dso, machine);if (dso__data (dso)->fd <0 ) { - ret = -errno;dso__data (dso)->status =DSO_DATA_STATUS_ERROR ; +/* fd is always negative here: -errno or -1, never 0 */ + ret =dso__data (dso)->fd; +assert (ret <0 );
[Severity: Low]
Is it possible to ensure <assert.h> is included for this usage as well?
Quoted 2 lines
tools/perf/util/dso.c +0
-0goto out; }
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=3
Arnaldo Carvalho de Melo
maintainerPATCH v4 4/5 perf dso: Guard against cache underflow on short reads in dso_cache__memcpy()
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
dso_cache__memcpy() computes cache_offset = offset - cache->offset, then cache_size = min(cache->size - cache_offset, size). The RB tree lookup in __dso_cache__find() matches using the full DSO__DATA_CACHE_SIZE window, but cache->size reflects the actual pread return value from dso_cache__populate().
A short pread (e.g. near end-of-file) makes cache->size smaller than DSO__DATA_CACHE_SIZE. If a subsequent access targets an offset past cache->offset + cache->size but within the DSO__DATA_CACHE_SIZE window, the cache entry is found but cache_offset exceeds cache->size. Since both are u64, the subtraction cache->size - cache_offset wraps to a large value, min() selects the caller's size, and memcpy reads out of bounds.
Return 0 for an offset past the valid cached data. For a regular file a short pread only happens at end-of-file, so 0 is what a direct pread() at that offset would return: cached_io() stops its read loop as on EOF. Re-reading from the backing file would not help — a second pread at the same offset returns the same short count.
--- tools/perf/util/dso.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-)
tools/perf/util/dso.c +14 -1
@@ -1005,7 +1005,20 @@ static ssize_t dso_cache__memcpy(struct dso_cache *cache, u64 offset, u8 *data, u64 size, bool out) {u64 cache_offset = offset - cache->offset; -u64 cache_size =min (cache->size - cache_offset, size); +u64 cache_size; + +/* +* The RB tree matches using DSO__DATA_CACHE_SIZE, but a short +* pread may leave cache->size smaller. For a regular file a +* short pread only happens at end-of-file, so an offset past +* the valid data is EOF: return 0, matching what a direct +* pread() at that offset would return, and cached_io() then +* stops its read loop. +*/ +if (cache_offset >= cache->size) +return 0 ; + + cache_size =min (cache->size - cache_offset, size);if (out)memcpy (data, cache->data + cache_offset, cache_size);
Signature · 1 line
2.55.0
[email protected]
AI reviewRe: [PATCH 4/5] perf dso: Guard against cache underflow on short reads in dso_cache__memcpy()
Sashiko has reviewed this patch and found no issues. It looks great!
Quoted 22 lines
From: Arnaldo Carvalho de Melo <[email protected]>
dso_cache__memcpy()computescache_offset= offset -cache->offset, thencache_size=min(cache->size - cache_offset, size). The RB tree lookup in__dso_cache__find()matches using the fullDSO__DATA_CACHE_SIZEwindow, butcache->sizereflects the actual pread return value fromdso_cache__populate().A short pread (e.g. near end-of-file) makes
cache->sizesmaller thanDSO__DATA_CACHE_SIZE. If a subsequent access targets an offset pastcache->offset+cache->sizebut within theDSO__DATA_CACHE_SIZEwindow, the cache entry is found butcache_offsetexceedscache->size. Since both are u64, the subtractioncache->size-cache_offsetwraps to a large value,min()selects the caller's size, and memcpy reads out of bounds.Return 0 for an offset past the valid cached data. For a regular file a short pread only happens at end-of-file, so 0 is what a direct
pread()at that offset would return:cached_io()stops its read loop as on EOF. Re-reading from the backing file would not help — a second pread at the same offset returns the same short count.
Quoted 5 lines
Fixes 366df72657e0 ("perf dso: Refactor dso_cache__read()")Reported-by sashiko-bot <[email protected]>Cc Adrian Hunter <[email protected]>Assisted-by Claude:claude-opus-4.6Signed-off-by Arnaldo Carvalho de Melo <[email protected]>
Sashiko has reviewed this patch and found no issues. It looks great!
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=4
Arnaldo Carvalho de Melo
maintainerPATCH v4 5/5 perf dso: Replace assert with runtime check in dso__read_symbol()
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
dso__read_symbol() asserts that len <= jited_prog_len, where len comes from sym->end - sym->start (parsed from PERF_RECORD_KSYMBOL in perf.data). Both values originate from untrusted file input.
With NDEBUG (production builds), the assert is compiled out, allowing an out-of-bounds heap read when the BPF program buffer is accessed. Without NDEBUG, a crafted perf.data crashes perf with an assertion failure.
Replace the assert with a runtime bounds check that returns NULL with an appropriate error code, matching the existing error handling pattern in this function.
--- tools/perf/util/dso.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
tools/perf/util/dso.c +6 -1
@@ -2033,7 +2033,12 @@ const u8 *dso__read_symbol(struct dso *dso, const char *symfs_filename, errno =SYMBOL_ANNOTATE_ERRNO__BPF_MISSING_BTF ; return NULL; } -assert (len <= info_linear->info.jited_prog_len); +if (len > info_linear->info.jited_prog_len) { +pr_debug ("BPF symbol length %zu exceeds jited_prog_len %u\n" , + len, info_linear->info.jited_prog_len); + errno =SYMBOL_ANNOTATE_ERRNO__BPF_MISSING_BTF ; +return NULL ; + } *out_buf_len = len;return (const u8 *)(uintptr_t )(info_linear->info.jited_prog_insns);#else
Signature · 1 line
2.55.0
[email protected]
AI reviewRe: [PATCH 5/5] perf dso: Replace assert with runtime check in dso__read_symbol()
Sashiko has reviewed this patch and found no issues. It looks great!
Quoted 15 lines
From: Arnaldo Carvalho de Melo <[email protected]>
dso__read_symbol()asserts that len <=jited_prog_len, where len comes fromsym->end-sym->start(parsed fromPERF_RECORD_KSYMBOLin perf.data). Both values originate from untrusted file input.With NDEBUG (production builds), the assert is compiled out, allowing an out-of-bounds heap read when the BPF program buffer is accessed. Without NDEBUG, a crafted perf.data crashes perf with an assertion failure.
Replace the assert with a runtime bounds check that returns NULL with an appropriate error code, matching the existing error handling pattern in this function.
Quoted 6 lines
Fixes aa04707f507e ("perf dso: Support BPF programs in dso__read_symbol()")Reported-by sashiko-bot <[email protected]>Cc Ian Rogers <[email protected]>Cc Song Liu <[email protected]>Assisted-by Claude:claude-opus-4.6Signed-off-by Arnaldo Carvalho de Melo <[email protected]>
Sashiko has reviewed this patch and found no issues. It looks great!
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=5