Harden and extend ELF build ID parsing logic

PATCH Superseded Andrii Nakryiko 3 participants 14 messages Jul 9, 2024 – Jul 10, 2024

Subsystems (10)

Memory management Kselftest Library code BPF +6 more subsystems

lore

Share & tools

lore

Andrii Nakryiko

author maintainer

2024-07-09 20:42 UTC #0 lore raw

cover Harden and extend ELF build ID parsing logic

The goal of this patch set is to extend existing ELF build ID parsing logic, currently mostly used by BPF subsystem, with support for working in sleepable mode in which memory faults are allowed and can be relied upon to fetch relevant parts of ELF file to find and fetch .note.gnu.build-id information.

This is useful and important for BPF subsystem itself, but also for PROCMAP_QUERY ioctl(), built atop of /proc/<pid>/maps functionality (see [0]), which makes use of the same build_id_parse() functionality. PROCMAP_QUERY is always called from sleepable user process context, so it doesn't have to suffer from current restrictions of build_id_parse() which are due to the NMI context assumption.

Along the way, we harden the logic to avoid TOCTOU problems. We also lift existing limitations of only working as long as ELF program headers and build ID note section is contained strictly within the very first page of ELF file.

We achieve all of the above without duplication of logic between sleepable and non-sleepable modes through freader abstraction that manages underlying page cache page (on demand) and giving a simple to use direct memory access interface. With that, single page restrictions and adding sleepable mode support is rather straightforward.

We also extend existing set of BPF selftests with a few tests targeting build ID logic across sleepable and non-sleepabe contexts (we utilize sleepable and non-sleepable uprobes for that).

[0] https://lore.kernel.org/linux-mm/[email protected]/

Andrii Nakryiko (10):
lib/buildid: add single page-based file reader abstraction lib/buildid: take into account e_phoff when fetching program headers lib/buildid: remove single-page limit for PHDR search lib/buildid: rename build_id_parse() into build_id_parse_nofault() lib/buildid: implement sleepable build_id_parse() API lib/buildid: don't limit .note.gnu.build-id to the first page in ELF lib/buildid: harden build ID parsing logic some more
bpf: decouple stack_map_get_build_id_offset() from
perf_callchain_entry
bpf: wire up sleepable bpf_get_stack() and bpf_get_task_stack()
helpers
selftests/bpf: add build ID tests

 include/linux/bpf.h                           |   2 +
 include/linux/buildid.h                       |   4 +-
 kernel/bpf/stackmap.c                         | 131 +++++--
 kernel/events/core.c                          |   2 +-
 kernel/trace/bpf_trace.c                      |   5 +-
 lib/buildid.c                                 | 370 +++++++++++++-----
 tools/testing/selftests/bpf/Makefile          |   5 +-
 .../selftests/bpf/prog_tests/build_id.c       | 118 ++++++
 .../selftests/bpf/progs/test_build_id.c       |  31 ++
 tools/testing/selftests/bpf/uprobe_multi.c    |  34 ++
 tools/testing/selftests/bpf/uprobe_multi.ld   |  11 +
 11 files changed, 584 insertions(+), 129 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/build_id.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_build_id.c
 create mode 100644 tools/testing/selftests/bpf/uprobe_multi.ld

Signature · 1 line

2.43.0

Andrii Nakryiko

author maintainer

2024-07-09 20:42 UTC #1 lore raw

PATCH 1/10 lib/buildid: add single page-based file reader abstraction

Add freader abstraction that transparently manages fetching and local mapping of the underlying file page(s) and provides a simple direct data access interface.

freader_fetch() is the only and single interface necessary. It accepts file offset and desired number of bytes that should be accessed, and will return a kernel mapped pointer that caller can use to dereference data up to requested size. Requested size can't be bigger than the size of the extra buffer provided during initialization (because, worst case, all requested data has to be copied into it, so it's better to flag wrongly sized buffer unconditionally, regardless if requested data range is crossing page boundaries or not).

If page is not paged in, or some of the conditions are not satisfied, NULL is returned and more detailed error code can be accessed through freader->err field. This approach makes the usage of freader_fetch() cleaner.

To accommodate accessing file data that crosses page boundaries, user has to provide an extra buffer that will be used to make a local copy, if necessary. This is done to maintain a simple linear pointer data access interface.

We switch existing build ID parsing logic to it, without changing or lifting any of the existing constraints, yet. This will be done separately.

Given existing code was written with the assumption that it's always working with a single (first) page of the underlying ELF file, logic passes direct pointers around, which doesn't really work well with freader approach and would be limiting when removing the single page limitation. So we adjust all the logic to work in terms of file offsets.

There is also a memory buffer-based version (freader_init_from_mem()) for cases when desired data is already available in kernel memory. This is used for parsing vmlinux's own build ID note. In this mode assumption is that provided data starts at "file offset" zero, which works great when parsing ELF notes sections, as all the parsing logic is relative to note section's start.

Signed-off-by Andrii Nakryiko <[email protected]>

---
 lib/buildid.c | 278 +++++++++++++++++++++++++++++++++++++++-----------
 1 file changed, 217 insertions(+), 61 deletions(-)

lib/buildid.c +217 ~~-61~~

index 7954dd92e36c..1442a2483a8b 100644
@@ -8,38 +8,174 @@
 
 #define BUILD_ID 3
 
+struct freader {
+	void *buf;
+	u32 buf_sz;
+	int err;
+	union {
+		struct {
+			struct address_space *mapping;
+			struct page *page;
+			void *page_addr;
+			u64 file_off;
+		};
+		struct {
+			const char *data;
+			u64 data_sz;
+		};
+	};
+};
+
+static void freader_init_from_file(struct freader *r, void *buf, u32 buf_sz,
+				   struct address_space *mapping)
+{
+	memset(r, 0, sizeof(*r));
+	r->buf = buf;
+	r->buf_sz = buf_sz;
+	r->mapping = mapping;
+}
+
+static void freader_init_from_mem(struct freader *r, const char *data, u64 data_sz)
+{
+	memset(r, 0, sizeof(*r));
+	r->data = data;
+	r->data_sz = data_sz;
+}
+
+static void freader_put_page(struct freader *r)
+{
+	if (!r->page)
+		return;
+	kunmap_local(r->page_addr);
+	put_page(r->page);
+	r->page = NULL;
+}
+
+static int freader_get_page(struct freader *r, u64 file_off)
+{
+	pgoff_t pg_off = file_off >> PAGE_SHIFT;
+
+	freader_put_page(r);
+
+	r->page = find_get_page(r->mapping, pg_off);
+	if (!r->page)
+		return -EFAULT;	/* page not mapped */
+
+	r->page_addr = kmap_local_page(r->page);
+	r->file_off = file_off & PAGE_MASK;
+
+	return 0;
+}
+
+static const void *freader_fetch(struct freader *r, u64 file_off, size_t sz)
+{
+	int err;
+
+	/* provided internal temporary buffer should be sized correctly */
+	if (WARN_ON(r->buf && sz > r->buf_sz)) {
+		r->err = -E2BIG;
+		return NULL;
+	}
+
+	if (unlikely(file_off + sz < file_off)) {
+		r->err = -EOVERFLOW;
+		return NULL;
+	}
+
+	/* working with memory buffer is much more straightforward */
+	if (!r->buf) {
+		if (file_off + sz > r->data_sz) {
+			r->err = -ERANGE;
+			return NULL;
+		}
+		return r->data + file_off;
+	}
+
+	/* check if we need to fetch a different page first */
+	if (!r->page || file_off < r->file_off || file_off >= r->file_off + PAGE_SIZE) {
+		err = freader_get_page(r, file_off);
+		if (err) {
+			r->err = err;
+			return NULL;
+		}
+	}
+
+	/* if requested data is crossing page boundaries, we have to copy
+	 * everything into our local buffer to keep a simple linear memory
+	 * access interface
+	 */
+	if (file_off + sz > r->file_off + PAGE_SIZE) {
+		int part_sz = r->file_off + PAGE_SIZE - file_off;
+
+		/* copy the part that resides in the current page */
+		memcpy(r->buf, r->page_addr + (file_off - r->file_off), part_sz);
+
+		/* fetch next page */
+		err = freader_get_page(r, r->file_off + PAGE_SIZE);
+		if (err) {
+			r->err = err;
+			return NULL;
+		}
+
+		/* copy the rest of requested data */
+		memcpy(r->buf + part_sz, r->page_addr, sz - part_sz);
+
+		return r->buf;
+	}
+
+	/* if data fits in a single page, just return direct pointer */
+	return r->page_addr + (file_off - r->file_off);
+}
+
+static void freader_cleanup(struct freader *r)
+{
+	freader_put_page(r);
+}
+
 /*
  * Parse build id from the note segment. This logic can be shared between
  * 32-bit and 64-bit system, because Elf32_Nhdr and Elf64_Nhdr are
  * identical.
  */
-static int parse_build_id_buf(unsigned char *build_id,
-			      __u32 *size,
-			      const void *note_start,
-			      Elf32_Word note_size)
+static int parse_build_id_buf(struct freader *r,
+			      unsigned char *build_id, __u32 *size,
+			      u64 note_offs, Elf32_Word note_size)
 {
-	Elf32_Word note_offs = 0, new_offs;
+	const char note_name[] = "GNU";
+	const size_t note_name_sz = sizeof(note_name);
+	u64 build_id_off, new_offs, note_end = note_offs + note_size;
+	u32 build_id_sz;
+	const Elf32_Nhdr *nhdr;
+	const char *data;
 
-	while (note_offs + sizeof(Elf32_Nhdr) < note_size) {
-		Elf32_Nhdr *nhdr = (Elf32_Nhdr *)(note_start + note_offs);
+	while (note_offs + sizeof(Elf32_Nhdr) < note_end) {
+		nhdr = freader_fetch(r, note_offs, sizeof(Elf32_Nhdr) + note_name_sz);
+		if (!nhdr)
+			return r->err;
 
 		if (nhdr->n_type == BUILD_ID &&
-		    nhdr->n_namesz == sizeof("GNU") &&
-		    !strcmp((char *)(nhdr + 1), "GNU") &&
+		    nhdr->n_namesz == note_name_sz &&
+		    !strcmp((char *)(nhdr + 1), note_name) &&
 		    nhdr->n_descsz > 0 &&
 		    nhdr->n_descsz <= BUILD_ID_SIZE_MAX) {
-			memcpy(build_id,
-			       note_start + note_offs +
-			       ALIGN(sizeof("GNU"), 4) + sizeof(Elf32_Nhdr),
-			       nhdr->n_descsz);
-			memset(build_id + nhdr->n_descsz, 0,
-			       BUILD_ID_SIZE_MAX - nhdr->n_descsz);
+
+			build_id_off = note_offs + sizeof(Elf32_Nhdr) + ALIGN(note_name_sz, 4);
+			build_id_sz = nhdr->n_descsz;
+
+			/* freader_fetch() will invalidate nhdr pointer */
+			data = freader_fetch(r, build_id_off, build_id_sz);
+			if (!data)
+				return r->err;
+
+			memcpy(build_id, data, build_id_sz);
+			memset(build_id + build_id_sz, 0, BUILD_ID_SIZE_MAX - build_id_sz);
 			if (size)
-				*size = nhdr->n_descsz;
+				*size = build_id_sz;
 			return 0;
 		}
+
 		new_offs = note_offs + sizeof(Elf32_Nhdr) +
-			ALIGN(nhdr->n_namesz, 4) + ALIGN(nhdr->n_descsz, 4);
+			   ALIGN(nhdr->n_namesz, 4) + ALIGN(nhdr->n_descsz, 4);
 		if (new_offs <= note_offs)  /* overflow */
 			break;
 		note_offs = new_offs;

@@ -48,73 +184,87 @@ static int parse_build_id_buf(unsigned char *build_id,
 	return -EINVAL;
 }
 
-static inline int parse_build_id(const void *page_addr,
+static inline int parse_build_id(struct freader *r,
 				 unsigned char *build_id,
 				 __u32 *size,
-				 const void *note_start,
+				 u64 note_start_off,
 				 Elf32_Word note_size)
 {
 	/* check for overflow */
-	if (note_start < page_addr || note_start + note_size < note_start)
+	if (note_start_off + note_size < note_start_off)
 		return -EINVAL;
 
 	/* only supports note that fits in the first page */
-	if (note_start + note_size > page_addr + PAGE_SIZE)
+	if (note_start_off + note_size > PAGE_SIZE)
 		return -EINVAL;
 
-	return parse_build_id_buf(build_id, size, note_start, note_size);
+	return parse_build_id_buf(r, build_id, size, note_start_off, note_size);
 }
 
 /* Parse build ID from 32-bit ELF */
-static int get_build_id_32(const void *page_addr, unsigned char *build_id,
-			   __u32 *size)
+static int get_build_id_32(struct freader *r, unsigned char *build_id, __u32 *size)
 {
-	Elf32_Ehdr *ehdr = (Elf32_Ehdr *)page_addr;
-	Elf32_Phdr *phdr;
-	int i;
+	const Elf32_Ehdr *ehdr;
+	const Elf32_Phdr *phdr;
+	__u32 phnum, i;
+
+	ehdr = freader_fetch(r, 0, sizeof(Elf32_Ehdr));
+	if (!ehdr)
+		return r->err;
+
+	/* subsequent freader_fetch() calls invalidate pointers, so remember locally */
+	phnum = ehdr->e_phnum;
 
 	/* only supports phdr that fits in one page */
-	if (ehdr->e_phnum >
-	    (PAGE_SIZE - sizeof(Elf32_Ehdr)) / sizeof(Elf32_Phdr))
+	if (phnum > (PAGE_SIZE - sizeof(Elf32_Ehdr)) / sizeof(Elf32_Phdr))
 		return -EINVAL;
 
-	phdr = (Elf32_Phdr *)(page_addr + sizeof(Elf32_Ehdr));
+	for (i = 0; i < phnum; ++i) {
+		phdr = freader_fetch(r, i * sizeof(Elf32_Phdr), sizeof(Elf32_Phdr));
+		if (!phdr)
+			return r->err;
 
-	for (i = 0; i < ehdr->e_phnum; ++i) {
-		if (phdr[i].p_type == PT_NOTE &&
-		    !parse_build_id(page_addr, build_id, size,
-				    page_addr + phdr[i].p_offset,
-				    phdr[i].p_filesz))
+		if (phdr->p_type == PT_NOTE &&
+		    !parse_build_id(r, build_id, size, phdr->p_offset, phdr->p_filesz))
 			return 0;
 	}
 	return -EINVAL;
 }
 
 /* Parse build ID from 64-bit ELF */
-static int get_build_id_64(const void *page_addr, unsigned char *build_id,
-			   __u32 *size)
+static int get_build_id_64(struct freader *r, unsigned char *build_id, __u32 *size)
 {
-	Elf64_Ehdr *ehdr = (Elf64_Ehdr *)page_addr;
-	Elf64_Phdr *phdr;
-	int i;
+	const Elf64_Ehdr *ehdr;
+	const Elf64_Phdr *phdr;
+	__u32 phnum, i;
+
+	ehdr = freader_fetch(r, 0, sizeof(Elf64_Ehdr));
+	if (!ehdr)
+		return r->err;
+
+	/* subsequent freader_fetch() calls invalidate pointers, so remember locally */
+	phnum = ehdr->e_phnum;
 
 	/* only supports phdr that fits in one page */
-	if (ehdr->e_phnum >
-	    (PAGE_SIZE - sizeof(Elf64_Ehdr)) / sizeof(Elf64_Phdr))
+	if (phnum > (PAGE_SIZE - sizeof(Elf64_Ehdr)) / sizeof(Elf64_Phdr))
 		return -EINVAL;
 
-	phdr = (Elf64_Phdr *)(page_addr + sizeof(Elf64_Ehdr));
+	for (i = 0; i < phnum; ++i) {
+		phdr = freader_fetch(r, i * sizeof(Elf64_Phdr), sizeof(Elf64_Phdr));
+		if (!phdr)
+			return r->err;
 
-	for (i = 0; i < ehdr->e_phnum; ++i) {
-		if (phdr[i].p_type == PT_NOTE &&
-		    !parse_build_id(page_addr, build_id, size,
-				    page_addr + phdr[i].p_offset,
-				    phdr[i].p_filesz))
+		if (phdr->p_type == PT_NOTE &&
+		    !parse_build_id(r, build_id, size, phdr->p_offset, phdr->p_filesz))
 			return 0;
 	}
+
 	return -EINVAL;
 }
 
+/* enough for Elf64_Ehdr, Elf64_Phdr, and all the smaller requests */
+#define MAX_FREADER_BUF_SZ 64
+
 /*
  * Parse build ID of ELF file mapped to vma
  * @vma:      vma object

@@ -126,22 +276,25 @@ static int get_build_id_64(const void *page_addr, unsigned char *build_id,
 int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
 		   __u32 *size)
 {
-	Elf32_Ehdr *ehdr;
-	struct page *page;
-	void *page_addr;
+	const Elf32_Ehdr *ehdr;
+	struct freader r;
+	char buf[MAX_FREADER_BUF_SZ];
 	int ret;
 
 	/* only works for page backed storage  */
 	if (!vma->vm_file)
 		return -EINVAL;
 
-	page = find_get_page(vma->vm_file->f_mapping, 0);
-	if (!page)
-		return -EFAULT;	/* page not mapped */
+	freader_init_from_file(&r, buf, sizeof(buf), vma->vm_file->f_mapping);
+
+	/* fetch first 18 bytes of ELF header for checks */
+	ehdr = freader_fetch(&r, 0, offsetofend(Elf32_Ehdr, e_type));
+	if (!ehdr) {
+		ret = r.err;
+		goto out;
+	}
 
 	ret = -EINVAL;
-	page_addr = kmap_local_page(page);
-	ehdr = (Elf32_Ehdr *)page_addr;
 
 	/* compare magic x7f "ELF" */
 	if (memcmp(ehdr->e_ident, ELFMAG, SELFMAG) != 0)

@@ -152,12 +305,11 @@ int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
 		goto out;
 
 	if (ehdr->e_ident[EI_CLASS] == ELFCLASS32)
-		ret = get_build_id_32(page_addr, build_id, size);
+		ret = get_build_id_32(&r, build_id, size);
 	else if (ehdr->e_ident[EI_CLASS] == ELFCLASS64)
-		ret = get_build_id_64(page_addr, build_id, size);
+		ret = get_build_id_64(&r, build_id, size);
 out:
-	kunmap_local(page_addr);
-	put_page(page);
+	freader_cleanup(&r);
 	return ret;
 }

@@ -171,7 +323,11 @@ int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
  */
 int build_id_parse_buf(const void *buf, unsigned char *build_id, u32 buf_size)
 {
-	return parse_build_id_buf(build_id, NULL, buf, buf_size);
+	struct freader r;
+
+	freader_init_from_mem(&r, buf, buf_size);
+
+	return parse_build_id_buf(&r, build_id, NULL, 0, buf_size);
 }
 
 #if IS_ENABLED(CONFIG_STACKTRACE_BUILD_ID) || IS_ENABLED(CONFIG_VMCORE_INFO)

Signature · 1 line

2.43.0

Andrii Nakryiko

author maintainer

2024-07-09 20:42 UTC #2 lore raw

PATCH 2/10 lib/buildid: take into account e_phoff when fetching program headers

In reply to Andrii Nakryiko

Current code assumption is that program (segment) headers are following ELF header immediately. This is a common case, but is not guaranteed. So take into account e_phoff field of the ELF header when accessing program headers.

Reported-by Alexey Dobriyan <[email protected]>

Signed-off-by Andrii Nakryiko <[email protected]>

---
 lib/buildid.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

lib/buildid.c +6 -3

index 1442a2483a8b..ce48ffab4111 100644
@@ -206,7 +206,7 @@ static int get_build_id_32(struct freader *r, unsigned char *build_id, __u32 *si
 {
 	const Elf32_Ehdr *ehdr;
 	const Elf32_Phdr *phdr;
-	__u32 phnum, i;
+	__u32 phnum, phoff, i;
 
 	ehdr = freader_fetch(r, 0, sizeof(Elf32_Ehdr));
 	if (!ehdr)

@@ -214,13 +214,14 @@ static int get_build_id_32(struct freader *r, unsigned char *build_id, __u32 *si
 
 	/* subsequent freader_fetch() calls invalidate pointers, so remember locally */
 	phnum = ehdr->e_phnum;
+	phoff = READ_ONCE(ehdr->e_phoff);
 
 	/* only supports phdr that fits in one page */
 	if (phnum > (PAGE_SIZE - sizeof(Elf32_Ehdr)) / sizeof(Elf32_Phdr))
 		return -EINVAL;
 
 	for (i = 0; i < phnum; ++i) {
-		phdr = freader_fetch(r, i * sizeof(Elf32_Phdr), sizeof(Elf32_Phdr));
+		phdr = freader_fetch(r, phoff + i * sizeof(Elf32_Phdr), sizeof(Elf32_Phdr));
 		if (!phdr)
 			return r->err;

@@ -237,6 +238,7 @@ static int get_build_id_64(struct freader *r, unsigned char *build_id, __u32 *si
 	const Elf64_Ehdr *ehdr;
 	const Elf64_Phdr *phdr;
 	__u32 phnum, i;
+	__u64 phoff;
 
 	ehdr = freader_fetch(r, 0, sizeof(Elf64_Ehdr));
 	if (!ehdr)

@@ -244,13 +246,14 @@ static int get_build_id_64(struct freader *r, unsigned char *build_id, __u32 *si
 
 	/* subsequent freader_fetch() calls invalidate pointers, so remember locally */
 	phnum = ehdr->e_phnum;
+	phoff = READ_ONCE(ehdr->e_phoff);
 
 	/* only supports phdr that fits in one page */
 	if (phnum > (PAGE_SIZE - sizeof(Elf64_Ehdr)) / sizeof(Elf64_Phdr))
 		return -EINVAL;
 
 	for (i = 0; i < phnum; ++i) {
-		phdr = freader_fetch(r, i * sizeof(Elf64_Phdr), sizeof(Elf64_Phdr));
+		phdr = freader_fetch(r, phoff + i * sizeof(Elf64_Phdr), sizeof(Elf64_Phdr));
 		if (!phdr)
 			return r->err;

Signature · 1 line

2.43.0

Andrii Nakryiko

author maintainer

2024-07-09 20:43 UTC #3 lore raw

PATCH 3/10 lib/buildid: remove single-page limit for PHDR search

In reply to Andrii Nakryiko

Now that freader allows to access multiple pages transparently, there is no need to limit program headers to the very first ELF file page. Remove this limitation, but still put some sane limit on amount of program headers that we are willing to iterate over (set arbitrarily to 256).

Signed-off-by Andrii Nakryiko <[email protected]>

---
 lib/buildid.c | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

lib/buildid.c +8 -6

index ce48ffab4111..49fcb9a549bf 100644
@@ -8,6 +8,8 @@
 
 #define BUILD_ID 3
 
+#define MAX_PHDR_CNT 256
+
 struct freader {
 	void *buf;
 	u32 buf_sz;

@@ -216,9 +218,9 @@ static int get_build_id_32(struct freader *r, unsigned char *build_id, __u32 *si
 	phnum = ehdr->e_phnum;
 	phoff = READ_ONCE(ehdr->e_phoff);
 
-	/* only supports phdr that fits in one page */
-	if (phnum > (PAGE_SIZE - sizeof(Elf32_Ehdr)) / sizeof(Elf32_Phdr))
-		return -EINVAL;
+	/* set upper bound on amount of segments (phdrs) we iterate */
+	if (phnum > MAX_PHDR_CNT)
+		phnum = MAX_PHDR_CNT;
 
 	for (i = 0; i < phnum; ++i) {
 		phdr = freader_fetch(r, phoff + i * sizeof(Elf32_Phdr), sizeof(Elf32_Phdr));

@@ -248,9 +250,9 @@ static int get_build_id_64(struct freader *r, unsigned char *build_id, __u32 *si
 	phnum = ehdr->e_phnum;
 	phoff = READ_ONCE(ehdr->e_phoff);
 
-	/* only supports phdr that fits in one page */
-	if (phnum > (PAGE_SIZE - sizeof(Elf64_Ehdr)) / sizeof(Elf64_Phdr))
-		return -EINVAL;
+	/* set upper bound on amount of segments (phdrs) we iterate */
+	if (phnum > MAX_PHDR_CNT)
+		phnum = MAX_PHDR_CNT;
 
 	for (i = 0; i < phnum; ++i) {
 		phdr = freader_fetch(r, phoff + i * sizeof(Elf64_Phdr), sizeof(Elf64_Phdr));

Signature · 1 line

2.43.0

Andrii Nakryiko

author maintainer

2024-07-09 20:43 UTC #4 lore raw

PATCH 4/10 lib/buildid: rename build_id_parse() into build_id_parse_nofault()

In reply to Andrii Nakryiko

Make it clear that build_id_parse() assumes that it can take no page fault by renaming it and current few users to build_id_parse_nofault().

Also add build_id_parse() stub, which will be implemented in subsequent patches, just to preserve succesful kernel compilation if another upcoming user of build_id_parse() (PROCMAP_QUERY ioctl() for /proc/<pid>/maps, see [0]) gets merged with bpf-next tree.

That ioctl() users of build_id_parse() doesn't have no-page-fault restriction, so it will automatically benefit from sleepable implementation.

[0] https://lore.kernel.org/linux-mm/[email protected]/

Signed-off-by Andrii Nakryiko <[email protected]>

---
 include/linux/buildid.h |  4 ++--
 kernel/bpf/stackmap.c   |  2 +-
 kernel/events/core.c    |  2 +-
 lib/buildid.c           | 24 +++++++++++++++++++++---
 4 files changed, 25 insertions(+), 7 deletions(-)

include/linux/buildid.h +2 -2

index 20aa3c2d89f7..014a88c41073 100644
@@ -7,8 +7,8 @@
 #define BUILD_ID_SIZE_MAX 20
 
 struct vm_area_struct;
-int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
-		   __u32 *size);
+int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id, __u32 *size);
+int build_id_parse_nofault(struct vm_area_struct *vma, unsigned char *build_id, __u32 *size);
 int build_id_parse_buf(const void *buf, unsigned char *build_id, u32 buf_size);
 
 #if IS_ENABLED(CONFIG_STACKTRACE_BUILD_ID) || IS_ENABLED(CONFIG_VMCORE_INFO)

kernel/bpf/stackmap.c +1 -1

index c99f8e5234ac..770ae8e88016 100644
@@ -156,7 +156,7 @@ static void stack_map_get_build_id_offset(struct bpf_stack_build_id *id_offs,
 			goto build_id_valid;
 		}
 		vma = find_vma(current->mm, ips[i]);
-		if (!vma || build_id_parse(vma, id_offs[i].build_id, NULL)) {
+		if (!vma || build_id_parse_nofault(vma, id_offs[i].build_id, NULL)) {
 			/* per entry fall back to ips */
 			id_offs[i].status = BPF_STACK_BUILD_ID_IP;
 			id_offs[i].ip = ips[i];

kernel/events/core.c +1 -1

index f0128c5ff278..e3927b688699 100644
@@ -8812,7 +8812,7 @@ static void perf_event_mmap_event(struct perf_mmap_event *mmap_event)
 	mmap_event->event_id.header.size = sizeof(mmap_event->event_id) + size;
 
 	if (atomic_read(&nr_build_id_events))
-		build_id_parse(vma, mmap_event->build_id, &mmap_event->build_id_size);
+		build_id_parse_nofault(vma, mmap_event->build_id, &mmap_event->build_id_size);
 
 	perf_iterate_sb(perf_event_mmap_output,
 		       mmap_event,

lib/buildid.c +21 -3

index 49fcb9a549bf..5f898fee43d7 100644
@@ -276,10 +276,12 @@ static int get_build_id_64(struct freader *r, unsigned char *build_id, __u32 *si
  * @build_id: buffer to store build id, at least BUILD_ID_SIZE long
  * @size:     returns actual build id size in case of success
  *
- * Return: 0 on success, -EINVAL otherwise
+ * Assumes no page fault can be taken, so if relevant portions of ELF file are
+ * not already paged in, fetching of build ID fails.
+ *
+ * Return: 0 on success; negative error, otherwise
  */
-int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
-		   __u32 *size)
+int build_id_parse_nofault(struct vm_area_struct *vma, unsigned char *build_id, __u32 *size)
 {
 	const Elf32_Ehdr *ehdr;
 	struct freader r;

@@ -318,6 +320,22 @@ int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
 	return ret;
 }
 
+/*
+ * Parse build ID of ELF file mapped to VMA
+ * @vma:      vma object
+ * @build_id: buffer to store build id, at least BUILD_ID_SIZE long
+ * @size:     returns actual build id size in case of success
+ *
+ * Assumes faultable context and can cause page faults to bring in file data
+ * into page cache.
+ *
+ * Return: 0 on success; negative error, otherwise
+ */
+int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id, __u32 *size)
+{
+	return -EOPNOTSUPP;
+}
+
 /**
  * build_id_parse_buf - Get build ID from a buffer
  * @buf:      ELF note section(s) to parse

Signature · 1 line

2.43.0

Andrii Nakryiko

author maintainer

2024-07-09 20:43 UTC #5 lore raw

PATCH 5/10 lib/buildid: implement sleepable build_id_parse() API

In reply to Andrii Nakryiko

Extend freader with a flag specifying whether it's OK to cause page fault to fetch file data that is not already physically present in memory. With this, it's now easy to wait for data if the caller is running in sleepable (faultable) context.

We utilize read_cache_folio() to bring the desired file page into page cache, after which the rest of the logic works just the same at page level.

Cc Shakeel Butt <[email protected]>

Cc Johannes Weiner <[email protected]>

Signed-off-by Andrii Nakryiko <[email protected]>

---
 lib/buildid.c | 49 ++++++++++++++++++++++++++++++++++---------------
 1 file changed, 34 insertions(+), 15 deletions(-)

lib/buildid.c +34 ~~-15~~

index 5f898fee43d7..23bfc811981a 100644
@@ -20,6 +20,7 @@ struct freader {
 			struct page *page;
 			void *page_addr;
 			u64 file_off;
+			bool may_fault;
 		};
 		struct {
 			const char *data;

@@ -29,12 +30,13 @@ struct freader {
 };
 
 static void freader_init_from_file(struct freader *r, void *buf, u32 buf_sz,
-				   struct address_space *mapping)
+				   struct address_space *mapping, bool may_fault)
 {
 	memset(r, 0, sizeof(*r));
 	r->buf = buf;
 	r->buf_sz = buf_sz;
 	r->mapping = mapping;
+	r->may_fault = may_fault;
 }
 
 static void freader_init_from_mem(struct freader *r, const char *data, u64 data_sz)

@@ -60,6 +62,17 @@ static int freader_get_page(struct freader *r, u64 file_off)
 	freader_put_page(r);
 
 	r->page = find_get_page(r->mapping, pg_off);
+
+	if (!r->page && r->may_fault) {
+		struct folio *folio;
+
+		folio = read_cache_folio(r->mapping, pg_off, NULL, NULL);
+		if (IS_ERR(folio))
+			return PTR_ERR(folio);
+
+		r->page = folio_file_page(folio, pg_off);
+	}
+
 	if (!r->page)
 		return -EFAULT;	/* page not mapped */

@@ -270,18 +283,8 @@ static int get_build_id_64(struct freader *r, unsigned char *build_id, __u32 *si
 /* enough for Elf64_Ehdr, Elf64_Phdr, and all the smaller requests */
 #define MAX_FREADER_BUF_SZ 64
 
-/*
- * Parse build ID of ELF file mapped to vma
- * @vma:      vma object
- * @build_id: buffer to store build id, at least BUILD_ID_SIZE long
- * @size:     returns actual build id size in case of success
- *
- * Assumes no page fault can be taken, so if relevant portions of ELF file are
- * not already paged in, fetching of build ID fails.
- *
- * Return: 0 on success; negative error, otherwise
- */
-int build_id_parse_nofault(struct vm_area_struct *vma, unsigned char *build_id, __u32 *size)
+static int __build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
+			    __u32 *size, bool may_fault)
 {
 	const Elf32_Ehdr *ehdr;
 	struct freader r;

@@ -292,7 +295,7 @@ int build_id_parse_nofault(struct vm_area_struct *vma, unsigned char *build_id,
 	if (!vma->vm_file)
 		return -EINVAL;
 
-	freader_init_from_file(&r, buf, sizeof(buf), vma->vm_file->f_mapping);
+	freader_init_from_file(&r, buf, sizeof(buf), vma->vm_file->f_mapping, may_fault);
 
 	/* fetch first 18 bytes of ELF header for checks */
 	ehdr = freader_fetch(&r, 0, offsetofend(Elf32_Ehdr, e_type));

@@ -320,6 +323,22 @@ int build_id_parse_nofault(struct vm_area_struct *vma, unsigned char *build_id,
 	return ret;
 }
 
+/*
+ * Parse build ID of ELF file mapped to vma
+ * @vma:      vma object
+ * @build_id: buffer to store build id, at least BUILD_ID_SIZE long
+ * @size:     returns actual build id size in case of success
+ *
+ * Assumes no page fault can be taken, so if relevant portions of ELF file are
+ * not already paged in, fetching of build ID fails.
+ *
+ * Return: 0 on success; negative error, otherwise
+ */
+int build_id_parse_nofault(struct vm_area_struct *vma, unsigned char *build_id, __u32 *size)
+{
+	return __build_id_parse(vma, build_id, size, false /* !may_fault */);
+}
+
 /*
  * Parse build ID of ELF file mapped to VMA
  * @vma:      vma object

@@ -333,7 +352,7 @@ int build_id_parse_nofault(struct vm_area_struct *vma, unsigned char *build_id,
  */
 int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id, __u32 *size)
 {
-	return -EOPNOTSUPP;
+	return __build_id_parse(vma, build_id, size, true /* may_fault */);
 }
 
 /**

Signature · 1 line

2.43.0

Andrii Nakryiko

author maintainer

2024-07-09 20:43 UTC #6 lore raw

PATCH 6/10 lib/buildid: don't limit .note.gnu.build-id to the first page in ELF

In reply to Andrii Nakryiko

With freader we don't need to restrict ourselves to a single page, so let's allow ELF notes to be at any valid position with the file.

We also merge parse_build_id() and parse_build_id_buf() as now the only difference between them is note offset overflow, which makes sense to check in all situations.

Signed-off-by Andrii Nakryiko <[email protected]>

---
 lib/buildid.c | 28 +++++++---------------------
 1 file changed, 7 insertions(+), 21 deletions(-)

lib/buildid.c +7 ~~-21~~

index 23bfc811981a..419966d88cd5 100644
@@ -152,9 +152,8 @@ static void freader_cleanup(struct freader *r)
  * 32-bit and 64-bit system, because Elf32_Nhdr and Elf64_Nhdr are
  * identical.
  */
-static int parse_build_id_buf(struct freader *r,
-			      unsigned char *build_id, __u32 *size,
-			      u64 note_offs, Elf32_Word note_size)
+static int parse_build_id(struct freader *r, unsigned char *build_id, __u32 *size,
+			  u64 note_offs, Elf32_Word note_size)
 {
 	const char note_name[] = "GNU";
 	const size_t note_name_sz = sizeof(note_name);

@@ -163,6 +162,10 @@ static int parse_build_id_buf(struct freader *r,
 	const Elf32_Nhdr *nhdr;
 	const char *data;
 
+	/* check for overflow */
+	if (note_offs + note_size < note_offs)
+		return -EINVAL;
+
 	while (note_offs + sizeof(Elf32_Nhdr) < note_end) {
 		nhdr = freader_fetch(r, note_offs, sizeof(Elf32_Nhdr) + note_name_sz);
 		if (!nhdr)

@@ -199,23 +202,6 @@ static int parse_build_id_buf(struct freader *r,
 	return -EINVAL;
 }
 
-static inline int parse_build_id(struct freader *r,
-				 unsigned char *build_id,
-				 __u32 *size,
-				 u64 note_start_off,
-				 Elf32_Word note_size)
-{
-	/* check for overflow */
-	if (note_start_off + note_size < note_start_off)
-		return -EINVAL;
-
-	/* only supports note that fits in the first page */
-	if (note_start_off + note_size > PAGE_SIZE)
-		return -EINVAL;
-
-	return parse_build_id_buf(r, build_id, size, note_start_off, note_size);
-}
-
 /* Parse build ID from 32-bit ELF */
 static int get_build_id_32(struct freader *r, unsigned char *build_id, __u32 *size)
 {

@@ -369,7 +355,7 @@ int build_id_parse_buf(const void *buf, unsigned char *build_id, u32 buf_size)
 
 	freader_init_from_mem(&r, buf, buf_size);
 
-	return parse_build_id_buf(&r, build_id, NULL, 0, buf_size);
+	return parse_build_id(&r, build_id, NULL, 0, buf_size);
 }
 
 #if IS_ENABLED(CONFIG_STACKTRACE_BUILD_ID) || IS_ENABLED(CONFIG_VMCORE_INFO)

Signature · 1 line

2.43.0

Andrii Nakryiko

author maintainer

2024-07-09 20:43 UTC #7 lore raw

PATCH 7/10 lib/buildid: harden build ID parsing logic some more

In reply to Andrii Nakryiko

Harden build ID parsing logic some more, adding explicit READ_ONCE() when fetching values that we then use to check correctness and various note iteration invariants.

Signed-off-by Andrii Nakryiko <[email protected]>

---
 lib/buildid.c | 32 +++++++++++++++++---------------
 1 file changed, 17 insertions(+), 15 deletions(-)

lib/buildid.c +17 ~~-15~~

index 419966d88cd5..7e36a32fbb90 100644
@@ -158,7 +158,7 @@ static int parse_build_id(struct freader *r, unsigned char *build_id, __u32 *siz
 	const char note_name[] = "GNU";
 	const size_t note_name_sz = sizeof(note_name);
 	u64 build_id_off, new_offs, note_end = note_offs + note_size;
-	u32 build_id_sz;
+	u32 build_id_sz, name_sz, desc_sz;
 	const Elf32_Nhdr *nhdr;
 	const char *data;

@@ -171,14 +171,15 @@ static int parse_build_id(struct freader *r, unsigned char *build_id, __u32 *siz
 		if (!nhdr)
 			return r->err;
 
-		if (nhdr->n_type == BUILD_ID &&
-		    nhdr->n_namesz == note_name_sz &&
-		    !strcmp((char *)(nhdr + 1), note_name) &&
-		    nhdr->n_descsz > 0 &&
-		    nhdr->n_descsz <= BUILD_ID_SIZE_MAX) {
+		name_sz = READ_ONCE(nhdr->n_namesz);
+		desc_sz = READ_ONCE(nhdr->n_descsz);
+		if (READ_ONCE(nhdr->n_type) == BUILD_ID &&
+		    name_sz == note_name_sz &&
+		    !strncmp((char *)(nhdr + 1), note_name, note_name_sz) &&
+		    desc_sz > 0 && desc_sz <= BUILD_ID_SIZE_MAX) {
 
 			build_id_off = note_offs + sizeof(Elf32_Nhdr) + ALIGN(note_name_sz, 4);
-			build_id_sz = nhdr->n_descsz;
+			build_id_sz = desc_sz;
 
 			/* freader_fetch() will invalidate nhdr pointer */
 			data = freader_fetch(r, build_id_off, build_id_sz);

@@ -192,8 +193,7 @@ static int parse_build_id(struct freader *r, unsigned char *build_id, __u32 *siz
 			return 0;
 		}
 
-		new_offs = note_offs + sizeof(Elf32_Nhdr) +
-			   ALIGN(nhdr->n_namesz, 4) + ALIGN(nhdr->n_descsz, 4);
+		new_offs = note_offs + sizeof(Elf32_Nhdr) + ALIGN(name_sz, 4) + ALIGN(desc_sz, 4);
 		if (new_offs <= note_offs)  /* overflow */
 			break;
 		note_offs = new_offs;

@@ -214,7 +214,7 @@ static int get_build_id_32(struct freader *r, unsigned char *build_id, __u32 *si
 		return r->err;
 
 	/* subsequent freader_fetch() calls invalidate pointers, so remember locally */
-	phnum = ehdr->e_phnum;
+	phnum = READ_ONCE(ehdr->e_phnum);
 	phoff = READ_ONCE(ehdr->e_phoff);
 
 	/* set upper bound on amount of segments (phdrs) we iterate */

@@ -226,8 +226,9 @@ static int get_build_id_32(struct freader *r, unsigned char *build_id, __u32 *si
 		if (!phdr)
 			return r->err;
 
-		if (phdr->p_type == PT_NOTE &&
-		    !parse_build_id(r, build_id, size, phdr->p_offset, phdr->p_filesz))
+		if (READ_ONCE(phdr->p_type) == PT_NOTE &&
+		    !parse_build_id(r, build_id, size,
+				    READ_ONCE(phdr->p_offset), READ_ONCE(phdr->p_filesz)))
 			return 0;
 	}
 	return -EINVAL;

@@ -246,7 +247,7 @@ static int get_build_id_64(struct freader *r, unsigned char *build_id, __u32 *si
 		return r->err;
 
 	/* subsequent freader_fetch() calls invalidate pointers, so remember locally */
-	phnum = ehdr->e_phnum;
+	phnum = READ_ONCE(ehdr->e_phnum);
 	phoff = READ_ONCE(ehdr->e_phoff);
 
 	/* set upper bound on amount of segments (phdrs) we iterate */

@@ -258,8 +259,9 @@ static int get_build_id_64(struct freader *r, unsigned char *build_id, __u32 *si
 		if (!phdr)
 			return r->err;
 
-		if (phdr->p_type == PT_NOTE &&
-		    !parse_build_id(r, build_id, size, phdr->p_offset, phdr->p_filesz))
+		if (READ_ONCE(phdr->p_type) == PT_NOTE &&
+		    !parse_build_id(r, build_id, size,
+				    READ_ONCE(phdr->p_offset), READ_ONCE(phdr->p_filesz)))
 			return 0;
 	}

Signature · 1 line

2.43.0

Andi Kleen

2024-07-10 20:45 UTC #8 lore raw

Re: [PATCH bpf-next 07/10] lib/buildid: harden build ID parsing logic some more

Quoted 3 lines · Andrii Nakryiko

Harden build ID parsing logic some more, adding explicit READ_ONCE() when fetching values that we then use to check correctness and various note iteration invariants.

Just sprinkling READ_ONCE all over doesn't necessarily fix the code. It is only needed for values that affect a loop or reference.

You have to fix stuff like this

static inline int parse_build_id(const void *page_addr,
unsigned char *build_id, __u32 *size, const void *note_start, Elf32_Word note_size)

{
        /* check for overflow */
        if (note_start < page_addr || note_start + note_size < note_start)
            ^^^^^^^^^^^^^^^^^^^^^^
                return -EINVAL;

which is C undefined (at least without -fwrapv-pointer) and can easily be miscompiled if it isn't already.

I suspect the code will need more work, especially since you're unwilling to consider any defense in depth measures.

-Andi

Andrii Nakryiko

2024-07-10 20:53 UTC #9 lore raw

Re: [PATCH bpf-next 07/10] lib/buildid: harden build ID parsing logic some more

Quoted 8 lines · Andi Kleen

Quoted 3 lines · Andrii Nakryiko
Harden build ID parsing logic some more, adding explicit READ_ONCE() when fetching values that we then use to check correctness and various note iteration invariants.
Just sprinkling READ_ONCE all over doesn't necessarily fix the code. It is only needed for values that affect a loop or reference.

Agreed, besides READ_ONCE(nhdr->n_type) == BUILD_ID and READ_ONCE(phdr->p_type) == PT_NOTE, which I added mostly just for consistency, the rest should be indeed read once and then checked, no? Do you see any other unnecessary READ_ONCE()s in this patch?

Quoted 8 lines

You have to fix stuff like this
static inline int parse_build_id(const void *page_addr,
unsigned char *build_id, __u32 *size, const void *note_start, Elf32_Word note_size)

Quoted 4 lines

{
        /* check for overflow */
        if (note_start < page_addr || note_start + note_size < note_start)
            ^^^^^^^^^^^^^^^^^^^^^^

this has been switched to u64-based offsets in patch #1, did you take a look at it?

Quoted 9 lines

                return -EINVAL;
which is C undefined (at least without -fwrapv-pointer) and can easily be miscompiled if it isn't already.
I suspect the code will need more work, especially since you're unwilling to consider any defense in depth measures.

Can you be a bit more specific about the remaining issues? I'm happy to fix whatever can and should be fixed (after the changes I already did in this patch set).

If by "defense in depth" you mean allowing this functionality only for executable VMAs, then yes, I refuse to do that, as I already explained.

Quoted 2 lines

-Andi

Andrii Nakryiko

author maintainer

2024-07-09 20:43 UTC #10 lore raw

PATCH 8/10 bpf: decouple stack_map_get_build_id_offset() from perf_callchain_entry

In reply to Andrii Nakryiko

Change stack_map_get_build_id_offset() which is used to convert stack trace IP addresses into build ID+offset pairs. Right now this function accepts an array of u64s as an input, and uses array of struct bpf_stack_build_id as an output.

This is problematic because u64 array is coming from perf_callchain_entry, which is (non-sleepable) RCU protected, so once we allows sleepable build ID fetching, this all breaks down.

But its actually pretty easy to make stack_map_get_build_id_offset() works with array of struct bpf_stack_build_id as both input and output. Which is what this patch is doing, eliminating the dependency on perf_callchain_entry. We require caller to fill out bpf_stack_build_id.ip fields (all other can be left uninitialized), and update in place as we do build ID resolution.

We make sure to READ_ONCE() and cache locally current IP value as we used it in a few places to find matching VMA and so on. Given this data is directly accessible and modifiable by user's BPF code, we should make sure to have a consistent view of it.

Signed-off-by Andrii Nakryiko <[email protected]>

---
 kernel/bpf/stackmap.c | 49 +++++++++++++++++++++++++++++--------------
 1 file changed, 33 insertions(+), 16 deletions(-)

kernel/bpf/stackmap.c +33 ~~-16~~

index 770ae8e88016..6457222b0b46 100644
@@ -124,8 +124,18 @@ static struct bpf_map *stack_map_alloc(union bpf_attr *attr)
 	return ERR_PTR(err);
 }
 
+/*
+ * Expects all id_offs[i].ip values to be set to correct initial IPs.
+ * They will be subsequently:
+ *   - either adjusted in place to a file offset, if build ID fetching
+ *     succeeds; in this case id_offs[i].build_id is set to correct build ID,
+ *     and id_offs[i].status is set to BPF_STACK_BUILD_ID_VALID;
+ *   - or IP will be kept intact, if build ID fetching failed; in this case
+ *     id_offs[i].build_id is zeroed out and id_offs[i].status is set to
+ *     BPF_STACK_BUILD_ID_IP.
+ */
 static void stack_map_get_build_id_offset(struct bpf_stack_build_id *id_offs,
-					  u64 *ips, u32 trace_nr, bool user)
+					  u32 trace_nr, bool user)
 {
 	int i;
 	struct mmap_unlock_irq_work *work = NULL;

@@ -142,30 +152,28 @@ static void stack_map_get_build_id_offset(struct bpf_stack_build_id *id_offs,
 		/* cannot access current->mm, fall back to ips */
 		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, BUILD_ID_SIZE_MAX);
 		}
 		return;
 	}
 
 	for (i = 0; i < trace_nr; i++) {
-		if (range_in_vma(prev_vma, ips[i], ips[i])) {
+		u64 ip = READ_ONCE(id_offs[i].ip);
+
+		if (range_in_vma(prev_vma, ip, ip)) {
 			vma = prev_vma;
-			memcpy(id_offs[i].build_id, prev_build_id,
-			       BUILD_ID_SIZE_MAX);
+			memcpy(id_offs[i].build_id, prev_build_id, BUILD_ID_SIZE_MAX);
 			goto build_id_valid;
 		}
-		vma = find_vma(current->mm, ips[i]);
+		vma = find_vma(current->mm, ip);
 		if (!vma || build_id_parse_nofault(vma, id_offs[i].build_id, NULL)) {
 			/* 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, BUILD_ID_SIZE_MAX);
 			continue;
 		}
 build_id_valid:
-		id_offs[i].offset = (vma->vm_pgoff << PAGE_SHIFT) + ips[i]
-			- vma->vm_start;
+		id_offs[i].offset = (vma->vm_pgoff << PAGE_SHIFT) + ip - vma->vm_start;
 		id_offs[i].status = BPF_STACK_BUILD_ID_VALID;
 		prev_vma = vma;
 		prev_build_id = id_offs[i].build_id;

@@ -216,7 +224,7 @@ static long __bpf_get_stackid(struct bpf_map *map,
 	struct bpf_stack_map *smap = container_of(map, struct bpf_stack_map, map);
 	struct stack_map_bucket *bucket, *new_bucket, *old_bucket;
 	u32 skip = flags & BPF_F_SKIP_FIELD_MASK;
-	u32 hash, id, trace_nr, trace_len;
+	u32 hash, id, trace_nr, trace_len, i;
 	bool user = flags & BPF_F_USER_STACK;
 	u64 *ips;
 	bool hash_matches;

@@ -238,15 +246,18 @@ static long __bpf_get_stackid(struct bpf_map *map,
 		return id;
 
 	if (stack_map_use_build_id(map)) {
+		struct bpf_stack_build_id *id_offs;
+
 		/* for build_id+offset, pop a bucket before slow cmp */
 		new_bucket = (struct stack_map_bucket *)
 			pcpu_freelist_pop(&smap->freelist);
 		if (unlikely(!new_bucket))
 			return -ENOMEM;
 		new_bucket->nr = trace_nr;
-		stack_map_get_build_id_offset(
-			(struct bpf_stack_build_id *)new_bucket->data,
-			ips, trace_nr, user);
+		id_offs = (struct bpf_stack_build_id *)new_bucket->data;
+		for (i = 0; i < trace_nr; i++)
+			id_offs[i].ip = ips[i];
+		stack_map_get_build_id_offset(id_offs, trace_nr, user);
 		trace_len = trace_nr * sizeof(struct bpf_stack_build_id);
 		if (hash_matches && bucket->nr == trace_nr &&
 		    memcmp(bucket->data, new_bucket->data, trace_len) == 0) {

@@ -445,10 +456,16 @@ static long __bpf_get_stack(struct pt_regs *regs, struct task_struct *task,
 	copy_len = trace_nr * elem_size;
 
 	ips = trace->ip + skip;
-	if (user && user_build_id)
-		stack_map_get_build_id_offset(buf, ips, trace_nr, user);
-	else
+	if (user && user_build_id) {
+		struct bpf_stack_build_id *id_offs = buf;
+		u32 i;
+
+		for (i = 0; i < trace_nr; i++)
+			id_offs[i].ip = ips[i];
+		stack_map_get_build_id_offset(buf, trace_nr, user);
+	} else {
 		memcpy(buf, ips, copy_len);
+	}
 
 	if (size > copy_len)
 		memset(buf + copy_len, 0, size - copy_len);

Signature · 1 line

2.43.0

Andrii Nakryiko

author maintainer

2024-07-09 20:43 UTC #11 lore raw

PATCH 9/10 bpf: wire up sleepable bpf_get_stack() and bpf_get_task_stack() helpers

In reply to Andrii Nakryiko

Add sleepable implementations of bpf_get_stack() and bpf_get_task_stack() helpers and allow them to be used from sleepable BPF program (e.g., sleepable uprobes).

Note, the stack trace IPs capturing itself is not sleepable (that would need to be a separate project), only build ID fetching is sleepable and thus more reliable, as it will wait for data to be paged in, if necessary. For that we make use of sleepable build_id_parse() implementation.

Now that build ID related internals in kernel/bpf/stackmap.c can be used both in sleepable and non-sleepable contexts, we need to add additional rcu_read_lock()/rcu_read_unlock() protection around fetching perf_callchain_entry, but with the refactoring in previous commit it's now pretty straightforward. We make sure to do rcu_read_unlock (in sleepable mode only) right before stack_map_get_build_id_offset() call which can sleep. By that time we don't have any more use of perf_callchain_entry.

Note, bpf_get_task_stack() will fail for user mode if task != current. And for kernel mode build ID are irrelevant. So in that sense adding sleepable bpf_get_task_stack() implementation is a no-op. It feel right to wire this up for symmetry and completeness, but I'm open to just dropping it until we support user && crosstask condition.

Signed-off-by Andrii Nakryiko <[email protected]>

---
 include/linux/bpf.h      |  2 +
 kernel/bpf/stackmap.c    | 90 ++++++++++++++++++++++++++++++++--------
 kernel/trace/bpf_trace.c |  5 ++-
 3 files changed, 77 insertions(+), 20 deletions(-)

include/linux/bpf.h +2 -0

index 960780ef04e1..1c28bc9a34e2 100644
@@ -3182,7 +3182,9 @@ extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
 extern const struct bpf_func_proto bpf_get_current_comm_proto;
 extern const struct bpf_func_proto bpf_get_stackid_proto;
 extern const struct bpf_func_proto bpf_get_stack_proto;
+extern const struct bpf_func_proto bpf_get_stack_sleepable_proto;
 extern const struct bpf_func_proto bpf_get_task_stack_proto;
+extern const struct bpf_func_proto bpf_get_task_stack_sleepable_proto;
 extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
 extern const struct bpf_func_proto bpf_get_stack_proto_pe;
 extern const struct bpf_func_proto bpf_sock_map_update_proto;

kernel/bpf/stackmap.c +72 ~~-18~~

202 lines not on this page · raw message

kernel/trace/bpf_trace.c +3 -2

index cd098846e251..c3845470f56d 100644
@@ -1598,7 +1598,8 @@ bpf_tracing_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 	case BPF_FUNC_jiffies64:
 		return &bpf_jiffies64_proto;
 	case BPF_FUNC_get_task_stack:
-		return &bpf_get_task_stack_proto;
+		return prog->sleepable ? &bpf_get_task_stack_sleepable_proto
+				       : &bpf_get_task_stack_proto;
 	case BPF_FUNC_copy_from_user:
 		return &bpf_copy_from_user_proto;
 	case BPF_FUNC_copy_from_user_task:

@@ -1654,7 +1655,7 @@ kprobe_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 	case BPF_FUNC_get_stackid:
 		return &bpf_get_stackid_proto;
 	case BPF_FUNC_get_stack:
-		return &bpf_get_stack_proto;
+		return prog->sleepable ? &bpf_get_stack_sleepable_proto : &bpf_get_stack_proto;
 #ifdef CONFIG_BPF_KPROBE_OVERRIDE
 	case BPF_FUNC_override_return:
 		return &bpf_override_return_proto;

Signature · 1 line

2.43.0

Andrii Nakryiko

author maintainer

2024-07-09 20:43 UTC #12 lore raw

PATCH 10/10 selftests/bpf: add build ID tests

In reply to Andrii Nakryiko

Add a new set of tests validating behavior of capturing stack traces with build ID. We extend uprobe_multi target binary with ability to trigger uprobe (so that we can capture stack traces from it), but also we allow to force build ID data to be either resident or non-resident in memory. That way we can validate that in non-sleepable context we won't get build ID (as expected), but with sleepable uprobes we will get that build ID regardless of it being physically present in memory.

Also, we add a small add-on linker script which reorders .note.gnu.build-id section and puts it after (big) .text section, putting build ID data outside of the very first page of ELF file. This will test all the relaxations we did in build ID parsing logic in kernel thanks to freader abstraction.

Signed-off-by Andrii Nakryiko <[email protected]>

---
 tools/testing/selftests/bpf/Makefile          |   5 +-
 .../selftests/bpf/prog_tests/build_id.c       | 118 ++++++++++++++++++
 .../selftests/bpf/progs/test_build_id.c       |  31 +++++
 tools/testing/selftests/bpf/uprobe_multi.c    |  34 +++++
 tools/testing/selftests/bpf/uprobe_multi.ld   |  11 ++
 5 files changed, 197 insertions(+), 2 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/build_id.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_build_id.c
 create mode 100644 tools/testing/selftests/bpf/uprobe_multi.ld

tools/testing/selftests/bpf/Makefile +3 -2

index e0b3887b3d2d..45f67e822f49 100644
@@ -762,9 +762,10 @@ $(OUTPUT)/veristat: $(OUTPUT)/veristat.o
 	$(call msg,BINARY,,$@)
 	$(Q)$(CC) $(CFLAGS) $(LDFLAGS) $(filter %.a %.o,$^) $(LDLIBS) -o $@
 
-$(OUTPUT)/uprobe_multi: uprobe_multi.c
+$(OUTPUT)/uprobe_multi: uprobe_multi.c uprobe_multi.ld
 	$(call msg,BINARY,,$@)
-	$(Q)$(CC) $(CFLAGS) -O0 $(LDFLAGS) $^ $(LDLIBS) -o $@
+	$(Q)$(CC) $(CFLAGS) -Wl,-T,uprobe_multi.ld -O0 $(LDFLAGS) 	\
+		$(filter-out %.ld,$^) $(LDLIBS) -o $@
 
 EXTRA_CLEAN := $(SCRATCH_DIR) $(HOST_SCRATCH_DIR)			\
 	prog_tests/tests.h map_tests/tests.h verifier/tests.h		\

tools/testing/selftests/bpf/prog_tests/build_id.c new file +118 -0

118 lines not on this page · raw message

tools/testing/selftests/bpf/progs/test_build_id.c new file +31 -0

new file mode 100644
index 000000000000..32ce59f9aa27
@@ -0,0 +1,31 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2024 Meta Platforms, Inc. and affiliates. */
+
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+
+struct bpf_stack_build_id stack_sleepable[128];
+int res_sleepable;
+
+struct bpf_stack_build_id stack_nofault[128];
+int res_nofault;
+
+SEC("uprobe.multi/./uprobe_multi:uprobe")
+int uprobe_nofault(struct pt_regs *ctx)
+{
+	res_nofault = bpf_get_stack(ctx, stack_nofault, sizeof(stack_nofault),
+				    BPF_F_USER_STACK | BPF_F_USER_BUILD_ID);
+
+	return 0;
+}
+
+SEC("uprobe.multi.s/./uprobe_multi:uprobe")
+int uprobe_sleepable(struct pt_regs *ctx)
+{
+	res_sleepable = bpf_get_stack(ctx, stack_sleepable, sizeof(stack_sleepable),
+				      BPF_F_USER_STACK | BPF_F_USER_BUILD_ID);
+
+	return 0;
+}
+
+char _license[] SEC("license") = "GPL";

tools/testing/selftests/bpf/uprobe_multi.c +34 -0

index 7ffa563ffeba..edd7ef3bd5cd 100644
@@ -2,6 +2,10 @@
 
 #include <stdio.h>
 #include <string.h>
+#include <stdbool.h>
+#include <stdint.h>
+#include <sys/mman.h>
+#include <unistd.h>
 #include <sdt.h>
 
 #define __PASTE(a, b) a##b

@@ -75,6 +79,32 @@ static int usdt(void)
 	return 0;
 }
 
+int __attribute__((weak)) uprobe(void)
+{
+	return 0;
+}
+
+extern char build_id_start[];
+extern char build_id_end[];
+
+int __attribute__((weak)) trigger_uprobe(bool build_id_resident)
+{
+	int page_sz = sysconf(_SC_PAGESIZE);
+	void *addr;
+
+	/* page-align build ID start */
+	addr = (void *)((uintptr_t)&build_id_start & ~(page_sz - 1));
+
+	if (build_id_resident)
+		madvise(addr, page_sz, MADV_POPULATE_READ);
+	else
+		madvise(addr, page_sz, MADV_PAGEOUT);
+
+	(void)uprobe();
+
+	return 0;
+}
+
 int main(int argc, char **argv)
 {
 	if (argc != 2)

@@ -84,6 +114,10 @@ int main(int argc, char **argv)
 		return bench();
 	if (!strcmp("usdt", argv[1]))
 		return usdt();
+	if (!strcmp("uprobe-paged-out", argv[1]))
+		return trigger_uprobe(false /* page-out build ID */);
+	if (!strcmp("uprobe-paged-in", argv[1]))
+		return trigger_uprobe(true /* page-in build ID */);
 
 error:
 	fprintf(stderr, "usage: %s <bench|usdt>\n", argv[0]);

tools/testing/selftests/bpf/uprobe_multi.ld new file +11 -0

new file mode 100644
index 000000000000..a2e94828bc8c
@@ -0,0 +1,11 @@
+SECTIONS
+{
+	. = ALIGN(4096);
+	.note.gnu.build-id : { *(.note.gnu.build-id) }
+	. = ALIGN(4096);
+}
+INSERT AFTER .text;
+
+build_id_start = ADDR(.note.gnu.build-id);
+build_id_end = ADDR(.note.gnu.build-id) + SIZEOF(.note.gnu.build-id);
+

Signature · 1 line

2.43.0

Andrii Nakryiko

2024-07-10 17:56 UTC #13 lore raw

Re: [PATCH bpf-next 10/10] selftests/bpf: add build ID tests

Quoted 16 lines · Andrii Nakryiko

Add a new set of tests validating behavior of capturing stack traces with build ID. We extend uprobe_multi target binary with ability to trigger uprobe (so that we can capture stack traces from it), but also we allow to force build ID data to be either resident or non-resident in memory. That way we can validate that in non-sleepable context we won't get build ID (as expected), but with sleepable uprobes we will get that build ID regardless of it being physically present in memory.
Also, we add a small add-on linker script which reorders .note.gnu.build-id section and puts it after (big) .text section, putting build ID data outside of the very first page of ELF file. This will test all the relaxations we did in build ID parsing logic in kernel thanks to freader abstraction.
Signed-off-by Andrii Nakryiko <[email protected]>

Quoted 11 lines

---
 tools/testing/selftests/bpf/Makefile          |   5 +-
 .../selftests/bpf/prog_tests/build_id.c       | 118 ++++++++++++++++++
 .../selftests/bpf/progs/test_build_id.c       |  31 +++++
 tools/testing/selftests/bpf/uprobe_multi.c    |  34 +++++
 tools/testing/selftests/bpf/uprobe_multi.ld   |  11 ++
 5 files changed, 197 insertions(+), 2 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/build_id.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_build_id.c
 create mode 100644 tools/testing/selftests/bpf/uprobe_multi.ld

FYI for anyone that decided to not look at this patch set because BPF CI build is failing. It's due to outdated libc headers that don't have MADV_POPULATE_READ constant. I've fixed it up with the usual #ifndef+#define+#endif block in selftests.

I'm not going to send another revision just to fix this up, I will wait for feedback. Thanks!

Quoted 8 lines

tools/testing/selftests/bpf/Makefile +0 -0

index e0b3887b3d2d..45f67e822f49 100644
@@ -762,9 +762,10 @@ $(OUTPUT)/veristat: $(OUTPUT)/veristat.o
        $(call msg,BINARY,,$@)
        $(Q)$(CC) $(CFLAGS) $(LDFLAGS) $(filter %.a %.o,$^) $(LDLIBS) -o $@

[...]