Skip to contentSkip to conversation

bpf: don't assume build-id length is always 20 bytes

PATCH Superseded Stanislav Fomichev 4 participants 11 messages Jan 15, 2019 – Jan 16, 2019

Subsystems (6)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/3 bpf: don't assume build-id length is always 20 bytes

Build-id length is not fixed to 20, it can be (man ld /--build-id):
* 128-bit (uuid)
* 160-bit (sha1)
* any length specified in ld --build-id=0xhexstring

To fix the issue of missing BPF_STACK_BUILD_ID_VALID for shorter build-ids, assume that build-id is somewhere in the range of 1 .. 20.
Set the remaining bytes to zero.

Fixes 615755a77b24 ("bpf: extend stackmap to save binary_build_id+offset instead of address")
---
 kernel/bpf/stackmap.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
kernel/bpf/stackmap.c +6 -2
index d9e2483669d0..8975d1768dcb 100644
@@ -180,11 +180,15 @@ static inline int stack_map_parse_build_id(void *page_addr,
 
 		if (nhdr->n_type == BPF_BUILD_ID &&
 		    nhdr->n_namesz == sizeof("GNU") &&
-		    nhdr->n_descsz == BPF_BUILD_ID_SIZE) {
+		    nhdr->n_descsz > 0 &&
+		    nhdr->n_descsz <= BPF_BUILD_ID_SIZE) {
+			__u32 len = min_t(__u32,
+					  BPF_BUILD_ID_SIZE, nhdr->n_descsz);
 			memcpy(build_id,
 			       note_start + note_offs +
 			       ALIGN(sizeof("GNU"), 4) + sizeof(Elf32_Nhdr),
-			       BPF_BUILD_ID_SIZE);
+			       len);
+			memset(build_id + len, 0, BPF_BUILD_ID_SIZE - len);
 			return 0;
 		}
 		new_offs = note_offs + sizeof(Elf32_Nhdr) +
Signature · 1 line
2.20.1.97.g81188d93c3-goog
#1 lore raw

PATCH 2/3 bpf: zero out build_id for BPF_STACK_BUILD_ID_IP

When returning BPF_STACK_BUILD_ID_IP from stack_map_get_build_id_offset, make sure that build_id field is empty. Since we are using percpu free list, there is a possibility that we might reuse some previous bpf_stack_build_id with non-zero build_id.

Fixes 615755a77b24 ("bpf: extend stackmap to save binary_build_id+offset instead of address")
---
 kernel/bpf/stackmap.c | 2 ++
 1 file changed, 2 insertions(+)
kernel/bpf/stackmap.c +2 -0
index 8975d1768dcb..f4b57c68c45f 100644
@@ -315,6 +315,7 @@ static void stack_map_get_build_id_offset(struct bpf_stack_build_id *id_offs,
 		for (i = 0; i < trace_nr; i++) {
 			id_offs[i].status = BPF_STACK_BUILD_ID_IP;
 			id_offs[i].ip = ips[i];
+			memset(id_offs[i].build_id, 0, BPF_BUILD_ID_SIZE);
 		}
 		return;
 	}
@@ -325,6 +326,7 @@ static void stack_map_get_build_id_offset(struct bpf_stack_build_id *id_offs,
 			/* per entry fall back to ips */
 			id_offs[i].status = BPF_STACK_BUILD_ID_IP;
 			id_offs[i].ip = ips[i];
+			memset(id_offs[i].build_id, 0, BPF_BUILD_ID_SIZE);
 			continue;
 		}
 		id_offs[i].offset = (vma->vm_pgoff << PAGE_SHIFT) + ips[i]
Signature · 1 line
2.20.1.97.g81188d93c3-goog
Acked-by on v1 2/3
#2 lore raw

Re: [PATCH bpf 2/3] bpf: zero out build_id for BPF_STACK_BUILD_ID_IP

Quoted 11 lines · Stanislav Fomichev

When returning BPF_STACK_BUILD_ID_IP from stack_map_get_build_id_offset, make sure that build_id field is empty. Since we are using percpu free list, there is a possibility that we might reuse some previous bpf_stack_build_id with non-zero build_id.

Fixes 615755a77b24 ("bpf: extend stackmap to save binary_build_id+offset instead of address")
Quoted 27 lines

---
kernel/bpf/stackmap.c | 2 ++

1 file changed, 2 insertions(+)
kernel/bpf/stackmap.c +2 -0
index 8975d1768dcb..f4b57c68c45f 100644
@@ -315,6 +315,7 @@ static void stack_map_get_build_id_offset(struct bpf_stack_build_id *id_offs,
 		for (i = 0; i < trace_nr; i++) {
 			id_offs[i].status = BPF_STACK_BUILD_ID_IP;
 			id_offs[i].ip = ips[i];
+			memset(id_offs[i].build_id, 0, BPF_BUILD_ID_SIZE);
 		}
 		return;
 	}
@@ -325,6 +326,7 @@ static void stack_map_get_build_id_offset(struct bpf_stack_build_id *id_offs,
 			/* per entry fall back to ips */
 			id_offs[i].status = BPF_STACK_BUILD_ID_IP;
 			id_offs[i].ip = ips[i];
+			memset(id_offs[i].build_id, 0, BPF_BUILD_ID_SIZE);
 			continue;
 		}
 		id_offs[i].offset = (vma->vm_pgoff << PAGE_SHIFT) + ips[i]
Signature · 1 line
2.20.1.97.g81188d93c3-goog
#3 lore raw

PATCH 3/3 selftests/bpf: retry tests that expect build-id

In reply to Stanislav Fomichev

While running test_progs in a loop I found out that I'm sometimes hitting "Didn't find expected build ID from the map" error.
Looking at stack_map_get_build_id_offset() it seems that it is racy (by design) and can sometimes return BPF_STACK_BUILD_ID_IP (i.e. can't trylock current->mm->mmap_sem).

Let's retry this test a single time.

Fixes 13790d1cc72c ("bpf: add selftest for stackmap with build_id in NMI context")
---
 tools/testing/selftests/bpf/test_progs.c | 30 ++++++++++++++++++++++++
 1 file changed, 30 insertions(+)
tools/testing/selftests/bpf/test_progs.c +30 -0
index 126fc624290d..25f0083a9b2e 100644
@@ -1188,7 +1188,9 @@ static void test_stacktrace_build_id(void)
 	int i, j;
 	struct bpf_stack_build_id id_offs[PERF_MAX_STACK_DEPTH];
 	int build_id_matches = 0;
+	int retry = 1;
 
+retry:
 	err = bpf_prog_load(file, BPF_PROG_TYPE_TRACEPOINT, &obj, &prog_fd);
 	if (CHECK(err, "prog_load", "err %d errno %d\n", err, errno))
 		goto out;
@@ -1301,6 +1303,19 @@ static void test_stacktrace_build_id(void)
 		previous_key = key;
 	} while (bpf_map_get_next_key(stackmap_fd, &previous_key, &key) == 0);
 
+	/* stack_map_get_build_id_offset() is racy and sometimes can return
+	 * BPF_STACK_BUILD_ID_IP instead of BPF_STACK_BUILD_ID_VALID;
+	 * try it one more time.
+	 */
+	if (build_id_matches < 1 && retry--) {
+		ioctl(pmu_fd, PERF_EVENT_IOC_DISABLE);
+		close(pmu_fd);
+		bpf_object__close(obj);
+		printf("%s:WARN:Didn't find expected build ID from the map, retrying\n",
+		       __func__);
+		goto retry;
+	}
+
 	if (CHECK(build_id_matches < 1, "build id match",
 		  "Didn't find expected build ID from the map\n"))
 		goto disable_pmu;
@@ -1341,7 +1356,9 @@ static void test_stacktrace_build_id_nmi(void)
 	int i, j;
 	struct bpf_stack_build_id id_offs[PERF_MAX_STACK_DEPTH];
 	int build_id_matches = 0;
+	int retry = 1;
 
+retry:
 	err = bpf_prog_load(file, BPF_PROG_TYPE_PERF_EVENT, &obj, &prog_fd);
 	if (CHECK(err, "prog_load", "err %d errno %d\n", err, errno))
 		return;
@@ -1436,6 +1453,19 @@ static void test_stacktrace_build_id_nmi(void)
 		previous_key = key;
 	} while (bpf_map_get_next_key(stackmap_fd, &previous_key, &key) == 0);
 
+	/* stack_map_get_build_id_offset() is racy and sometimes can return
+	 * BPF_STACK_BUILD_ID_IP instead of BPF_STACK_BUILD_ID_VALID;
+	 * try it one more time.
+	 */
+	if (build_id_matches < 1 && retry--) {
+		ioctl(pmu_fd, PERF_EVENT_IOC_DISABLE);
+		close(pmu_fd);
+		bpf_object__close(obj);
+		printf("%s:WARN:Didn't find expected build ID from the map, retrying\n",
+		       __func__);
+		goto retry;
+	}
+
 	if (CHECK(build_id_matches < 1, "build id match",
 		  "Didn't find expected build ID from the map\n"))
 		goto disable_pmu;
Signature · 1 line
2.20.1.97.g81188d93c3-goog
Acked-by on v1 3/3
#4 lore raw

Re: [PATCH bpf 3/3] selftests/bpf: retry tests that expect build-id

Quoted 10 lines · Stanislav Fomichev

While running test_progs in a loop I found out that I'm sometimes hitting "Didn't find expected build ID from the map" error.
Looking at stack_map_get_build_id_offset() it seems that it is racy (by design) and can sometimes return BPF_STACK_BUILD_ID_IP (i.e. can't trylock current->mm->mmap_sem).

Quoted 4 lines

Let's retry this test a single time.

Fixes 13790d1cc72c ("bpf: add selftest for stackmap with build_id in NMI context")
Quoted 71 lines

---
tools/testing/selftests/bpf/test_progs.c | 30 ++++++++++++++++++++++++

1 file changed, 30 insertions(+)
tools/testing/selftests/bpf/test_progs.c +30 -0
index 126fc624290d..25f0083a9b2e 100644
@@ -1188,7 +1188,9 @@ static void test_stacktrace_build_id(void)
 	int i, j;
 	struct bpf_stack_build_id id_offs[PERF_MAX_STACK_DEPTH];
 	int build_id_matches = 0;
+	int retry = 1;
 
+retry:
 	err = bpf_prog_load(file, BPF_PROG_TYPE_TRACEPOINT, &obj, &prog_fd);
 	if (CHECK(err, "prog_load", "err %d errno %d\n", err, errno))
 		goto out;
@@ -1301,6 +1303,19 @@ static void test_stacktrace_build_id(void)
 		previous_key = key;
 	} while (bpf_map_get_next_key(stackmap_fd, &previous_key, &key) == 0);
 
+	/* stack_map_get_build_id_offset() is racy and sometimes can return
+	 * BPF_STACK_BUILD_ID_IP instead of BPF_STACK_BUILD_ID_VALID;
+	 * try it one more time.
+	 */
+	if (build_id_matches < 1 && retry--) {
+		ioctl(pmu_fd, PERF_EVENT_IOC_DISABLE);
+		close(pmu_fd);
+		bpf_object__close(obj);
+		printf("%s:WARN:Didn't find expected build ID from the map, retrying\n",
+		       __func__);
+		goto retry;
+	}
+
 	if (CHECK(build_id_matches < 1, "build id match",
 		  "Didn't find expected build ID from the map\n"))
 		goto disable_pmu;
@@ -1341,7 +1356,9 @@ static void test_stacktrace_build_id_nmi(void)
 	int i, j;
 	struct bpf_stack_build_id id_offs[PERF_MAX_STACK_DEPTH];
 	int build_id_matches = 0;
+	int retry = 1;
 
+retry:
 	err = bpf_prog_load(file, BPF_PROG_TYPE_PERF_EVENT, &obj, &prog_fd);
 	if (CHECK(err, "prog_load", "err %d errno %d\n", err, errno))
 		return;
@@ -1436,6 +1453,19 @@ static void test_stacktrace_build_id_nmi(void)
 		previous_key = key;
 	} while (bpf_map_get_next_key(stackmap_fd, &previous_key, &key) == 0);
 
+	/* stack_map_get_build_id_offset() is racy and sometimes can return
+	 * BPF_STACK_BUILD_ID_IP instead of BPF_STACK_BUILD_ID_VALID;
+	 * try it one more time.
+	 */
+	if (build_id_matches < 1 && retry--) {
+		ioctl(pmu_fd, PERF_EVENT_IOC_DISABLE);
+		close(pmu_fd);
+		bpf_object__close(obj);
+		printf("%s:WARN:Didn't find expected build ID from the map, retrying\n",
+		       __func__);
+		goto retry;
+	}
+
 	if (CHECK(build_id_matches < 1, "build id match",
 		  "Didn't find expected build ID from the map\n"))
 		goto disable_pmu;
Signature · 1 line
2.20.1.97.g81188d93c3-goog
Acked-by on v1 1/3
#5 lore raw
In reply to Stanislav Fomichev
Quoted 24 lines · Stanislav Fomichev

Build-id length is not fixed to 20, it can be (man ld /--build-id):
* 128-bit (uuid)
* 160-bit (sha1)
* any length specified in ld --build-id=0xhexstring

To fix the issue of missing BPF_STACK_BUILD_ID_VALID for shorter build-ids, assume that build-id is somewhere in the range of 1 .. 20.
Set the remaining bytes to zero.

Fixes 615755a77b24 ("bpf: extend stackmap to save binary_build_id+offset instead of address")

---
kernel/bpf/stackmap.c | 8 ++++++--

1 file changed, 6 insertions(+), 2 deletions(-)
kernel/bpf/stackmap.c +0 -0
index d9e2483669d0..8975d1768dcb 100644

kernel/bpf/stackmap.c static inline int stack_map_parse_build_id(void *page_addr, · line 186

Quoted 8 lines
kernel/bpf/stackmap.c +4 -1
@@ -180,11 +180,15 @@ static inline int stack_map_parse_build_id(void *page_addr,
 
 		if (nhdr->n_type == BPF_BUILD_ID &&
 		    nhdr->n_namesz == sizeof("GNU") &&
-		    nhdr->n_descsz == BPF_BUILD_ID_SIZE) {
+		    nhdr->n_descsz > 0 &&
+		    nhdr->n_descsz <= BPF_BUILD_ID_SIZE) {
+			__u32 len = min_t(__u32,
+					  BPF_BUILD_ID_SIZE, nhdr->n_descsz);

Given the check above, we only need len = nhdr->n_descsz, right?

Other than this,

Thanks for the fix!

Quoted 12 lines
kernel/bpf/stackmap.c +2 -1
 			memcpy(build_id,
 			       note_start + note_offs +
 			       ALIGN(sizeof("GNU"), 4) + sizeof(Elf32_Nhdr),
-			       BPF_BUILD_ID_SIZE);
+			       len);
+			memset(build_id + len, 0, BPF_BUILD_ID_SIZE - len);
 			return 0;
 		}
 		new_offs = note_offs + sizeof(Elf32_Nhdr) +
Signature · 1 line
2.20.1.97.g81188d93c3-goog
#6 lore raw
Quoted 26 lines · Song Liu
Quoted 26 lines · Stanislav Fomichev

Build-id length is not fixed to 20, it can be (man ld /--build-id):
* 128-bit (uuid)
* 160-bit (sha1)
* any length specified in ld --build-id=0xhexstring

To fix the issue of missing BPF_STACK_BUILD_ID_VALID for shorter build-ids, assume that build-id is somewhere in the range of 1 .. 20.
Set the remaining bytes to zero.

Fixes 615755a77b24 ("bpf: extend stackmap to save binary_build_id+offset instead of address")

---
kernel/bpf/stackmap.c | 8 ++++++--

1 file changed, 6 insertions(+), 2 deletions(-)
kernel/bpf/stackmap.c +0 -0
index d9e2483669d0..8975d1768dcb 100644
@@ -180,11 +180,15 @@ static inline int stack_map_parse_build_id(void *page_addr,
 
Quoted 9 lines

kernel/bpf/stackmap.c static inline int stack_map_parse_build_id(void *page_addr, · line 186

Quoted 7 lines
kernel/bpf/stackmap.c +4 -1
              if (nhdr->n_type == BPF_BUILD_ID &&
                  nhdr->n_namesz == sizeof("GNU") &&
-                 nhdr->n_descsz == BPF_BUILD_ID_SIZE) {
+                 nhdr->n_descsz > 0 &&
+                 nhdr->n_descsz <= BPF_BUILD_ID_SIZE) {
+                     __u32 len = min_t(__u32,
+                                       BPF_BUILD_ID_SIZE, nhdr->n_descsz);

Given the check above, we only need len = nhdr->n_descsz, right?

Ah, correct, I'll fix in v2.
I initially had without if (nhdr->n_descsz <= BPF_BUILD_ID_SIZE) and clamped it here, but then decided that clamping is probably bad as well.

Quoted 20 lines

Other than this,

Thanks for the fix!

Quoted 12 lines
kernel/bpf/stackmap.c +2 -2
                      memcpy(build_id,
                             note_start + note_offs +
                             ALIGN(sizeof("GNU"), 4) + sizeof(Elf32_Nhdr),
-                            BPF_BUILD_ID_SIZE);
+                            len);
+                     memset(build_id + len, 0, BPF_BUILD_ID_SIZE - len);
                      return 0;
              }
              new_offs = note_offs + sizeof(Elf32_Nhdr) +
--

2.20.1.97.g81188d93c3-goog

#7 lore raw

PATCH v2 1/3 bpf: don't assume build-id length is always 20 bytes

In reply to Song Liu

Build-id length is not fixed to 20, it can be (man ld /--build-id):
* 128-bit (uuid)
* 160-bit (sha1)
* any length specified in ld --build-id=0xhexstring

To fix the issue of missing BPF_STACK_BUILD_ID_VALID for shorter build-ids, assume that build-id is somewhere in the range of 1 .. 20.
Set the remaining bytes to zero.

v2:
* don't introduce new "len = min(BPF_BUILD_ID_SIZE, nhdr->n_descsz)", we already know that nhdr->n_descsz <= BPF_BUILD_ID_SIZE if we enter this 'if' condition

Fixes 615755a77b24 ("bpf: extend stackmap to save binary_build_id+offset instead of address")
---
 kernel/bpf/stackmap.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
kernel/bpf/stackmap.c +5 -2
index d9e2483669d0..f9df545e92f6 100644
@@ -180,11 +180,14 @@ static inline int stack_map_parse_build_id(void *page_addr,
 
 		if (nhdr->n_type == BPF_BUILD_ID &&
 		    nhdr->n_namesz == sizeof("GNU") &&
-		    nhdr->n_descsz == BPF_BUILD_ID_SIZE) {
+		    nhdr->n_descsz > 0 &&
+		    nhdr->n_descsz <= BPF_BUILD_ID_SIZE) {
 			memcpy(build_id,
 			       note_start + note_offs +
 			       ALIGN(sizeof("GNU"), 4) + sizeof(Elf32_Nhdr),
-			       BPF_BUILD_ID_SIZE);
+			       nhdr->n_descsz);
+			memset(build_id + nhdr->n_descsz, 0,
+			       BPF_BUILD_ID_SIZE - nhdr->n_descsz);
 			return 0;
 		}
 		new_offs = note_offs + sizeof(Elf32_Nhdr) +
Signature · 1 line
2.20.1.97.g81188d93c3-goog
Acked-by on v2 1/3
#8 lore raw
Quoted 13 lines · Stanislav Fomichev

Build-id length is not fixed to 20, it can be (man ld /--build-id):
* 128-bit (uuid)
* 160-bit (sha1)
* any length specified in ld --build-id=0xhexstring

To fix the issue of missing BPF_STACK_BUILD_ID_VALID for shorter build-ids, assume that build-id is somewhere in the range of 1 .. 20.
Set the remaining bytes to zero.

Quoted 7 lines

v2:
* don't introduce new "len = min(BPF_BUILD_ID_SIZE, nhdr->n_descsz)", we already know that nhdr->n_descsz <= BPF_BUILD_ID_SIZE if we enter this 'if' condition

Fixes 615755a77b24 ("bpf: extend stackmap to save binary_build_id+offset instead of address")
Quoted 28 lines

---
kernel/bpf/stackmap.c | 7 +++++--

1 file changed, 5 insertions(+), 2 deletions(-)
kernel/bpf/stackmap.c +5 -2
index d9e2483669d0..f9df545e92f6 100644
@@ -180,11 +180,14 @@ static inline int stack_map_parse_build_id(void *page_addr,
 
 		if (nhdr->n_type == BPF_BUILD_ID &&
 		    nhdr->n_namesz == sizeof("GNU") &&
-		    nhdr->n_descsz == BPF_BUILD_ID_SIZE) {
+		    nhdr->n_descsz > 0 &&
+		    nhdr->n_descsz <= BPF_BUILD_ID_SIZE) {
 			memcpy(build_id,
 			       note_start + note_offs +
 			       ALIGN(sizeof("GNU"), 4) + sizeof(Elf32_Nhdr),
-			       BPF_BUILD_ID_SIZE);
+			       nhdr->n_descsz);
+			memset(build_id + nhdr->n_descsz, 0,
+			       BPF_BUILD_ID_SIZE - nhdr->n_descsz);
 			return 0;
 		}
 		new_offs = note_offs + sizeof(Elf32_Nhdr) +
Signature · 1 line
2.20.1.97.g81188d93c3-goog
maintainer
#9 lore raw
In reply to Stanislav Fomichev
Quoted 9 lines · Stanislav Fomichev

Build-id length is not fixed to 20, it can be (man ld /--build-id):
* 128-bit (uuid)
* 160-bit (sha1)
* any length specified in ld --build-id=0xhexstring

To fix the issue of missing BPF_STACK_BUILD_ID_VALID for shorter build-ids, assume that build-id is somewhere in the range of 1 .. 20.
Set the remaining bytes to zero.

Quoted 7 lines

v2:
* don't introduce new "len = min(BPF_BUILD_ID_SIZE, nhdr->n_descsz)", we already know that nhdr->n_descsz <= BPF_BUILD_ID_SIZE if we enter this 'if' condition

Fixes 615755a77b24 ("bpf: extend stackmap to save binary_build_id+offset instead of address")

Hmm, looks like rest of the v2 series didn't make it to the list. Please double check; just in case simply resend the full v2 set so it properly ends up in patchwork.

Thanks,
Daniel

#10 lore raw
Quoted 15 lines · Daniel Borkmann
Quoted 14 lines · Stanislav Fomichev

Build-id length is not fixed to 20, it can be (man ld /--build-id):
* 128-bit (uuid)
* 160-bit (sha1)
* any length specified in ld --build-id=0xhexstring

To fix the issue of missing BPF_STACK_BUILD_ID_VALID for shorter build-ids, assume that build-id is somewhere in the range of 1 .. 20.
Set the remaining bytes to zero.

v2:
* don't introduce new "len = min(BPF_BUILD_ID_SIZE, nhdr->n_descsz)", we already know that nhdr->n_descsz <= BPF_BUILD_ID_SIZE if we enter this 'if' condition

Quoted 6 lines
Quoted 2 lines
Fixes 615755a77b24 ("bpf: extend stackmap to save binary_build_id+offset instead of address")

Hmm, looks like rest of the v2 series didn't make it to the list. Please double check; just in case simply resend the full v2 set so it properly ends up in patchwork.

Oh, I didn't send the first two patches because I didn't change them, I'll resend full v2 series in a moment.

Quoted 3 lines

Thanks,
Daniel