Skip to contentSkip to conversation

BPF: Fix potential bad pointer dereference in bpf_sys_bpf()

PATCH Applied Jinghao Jia 3 participants 3 messages Jul 29, 2022 – Aug 5, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 BPF: Fix potential bad pointer dereference in bpf_sys_bpf()

The bpf_sys_bpf() helper function allows an eBPF program to load another eBPF program from within the kernel. In this case the argument union bpf_attr pointer (as well as the insns and license pointers inside) is a kernel address instead of a userspace address (which is the case of a usual bpf() syscall). To make the memory copying process in the syscall work in both cases, bpfptr_t was introduced to wrap around the pointer and distinguish its origin. Specifically, when copying memory contents from a bpfptr_t, a copy_from_user() is performed in case of a userspace address and a memcpy() is performed for a kernel address.

This can lead to problems because the in-kernel pointer is never checked for validity. The problem happens when an eBPF syscall program tries to call bpf_sys_bpf() to load a program but provides a bad insns pointer -- say 0xdeadbeef -- in the bpf_attr union. The helper calls __sys_bpf() which would then call bpf_prog_load() to load the program. bpf_prog_load() is responsible for copying the eBPF instructions to the newly allocated memory for the program; it creates a kernel bpfptr_t for insns and invokes copy_from_bpfptr(). Internally, all bpfptr_t operations are backed by the corresponding sockptr_t operations, which performs direct memcpy() on kernel pointers for copy_from/strncpy_from operations. Therefore, the code is always happy to dereference the bad pointer to trigger a un-handle-able page fault and in turn an oops. However, this is not supposed to happen because at that point the eBPF program is already verified and should not cause a memory error.

Sample KASAN trace:

[   25.685056][  T228] ==================================================================
[   25.685680][  T228] BUG: KASAN: user-memory-access in copy_from_bpfptr+0x21/0x30
[   25.686210][  T228] Read of size 80 at addr 00000000deadbeef by task poc/228
[   25.686732][  T228]
[   25.686893][  T228] CPU: 3 PID: 228 Comm: poc Not tainted 5.19.0-rc7 #7
[   25.687375][  T228] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS d55cb5a 04/01/2014
[   25.687991][  T228] Call Trace:
[   25.688223][  T228]  <TASK>
[   25.688429][  T228]  dump_stack_lvl+0x73/0x9e
[   25.688747][  T228]  print_report+0xea/0x200
[   25.689061][  T228]  ? copy_from_bpfptr+0x21/0x30
[   25.689401][  T228]  ? _printk+0x54/0x6e
[   25.689693][  T228]  ? _raw_spin_lock_irqsave+0x70/0xd0
[   25.690071][  T228]  ? copy_from_bpfptr+0x21/0x30
[   25.690412][  T228]  kasan_report+0xb5/0xe0
[   25.690716][  T228]  ? copy_from_bpfptr+0x21/0x30
[   25.691059][  T228]  kasan_check_range+0x2bd/0x2e0
[   25.691405][  T228]  ? copy_from_bpfptr+0x21/0x30
[   25.691734][  T228]  memcpy+0x25/0x60
[   25.692000][  T228]  copy_from_bpfptr+0x21/0x30
[   25.692328][  T228]  bpf_prog_load+0x604/0x9e0
[   25.692653][  T228]  ? cap_capable+0xb4/0xe0
[   25.692956][  T228]  ? security_capable+0x4f/0x70
[   25.693324][  T228]  __sys_bpf+0x3af/0x580
[   25.693635][  T228]  bpf_sys_bpf+0x45/0x240
[   25.693937][  T228]  bpf_prog_f0ec79a5a3caca46_bpf_func1+0xa2/0xbd
[   25.694394][  T228]  bpf_prog_run_pin_on_cpu+0x2f/0xb0
[   25.694756][  T228]  bpf_prog_test_run_syscall+0x146/0x1c0
[   25.695144][  T228]  bpf_prog_test_run+0x172/0x190
[   25.695487][  T228]  __sys_bpf+0x2c5/0x580
[   25.695776][  T228]  __x64_sys_bpf+0x3a/0x50
[   25.696084][  T228]  do_syscall_64+0x60/0x90
[   25.696393][  T228]  ? fpregs_assert_state_consistent+0x50/0x60
[   25.696815][  T228]  ? exit_to_user_mode_prepare+0x36/0xa0
[   25.697202][  T228]  ? syscall_exit_to_user_mode+0x20/0x40
[   25.697586][  T228]  ? do_syscall_64+0x6e/0x90
[   25.697899][  T228]  entry_SYSCALL_64_after_hwframe+0x63/0xcd
[   25.698312][  T228] RIP: 0033:0x7f6d543fb759
[   25.698624][  T228] Code: 08 5b 89 e8 5d c3 66 2e 0f 1f 84 00 00 00 00 00 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 97 a6 0e 00 f7 d8 64 89 01 48
[   25.699946][  T228] RSP: 002b:00007ffc3df78468 EFLAGS: 00000287 ORIG_RAX: 0000000000000141
[   25.700526][  T228] RAX: ffffffffffffffda RBX: 00007ffc3df78628 RCX: 00007f6d543fb759
[   25.701071][  T228] RDX: 0000000000000090 RSI: 00007ffc3df78478 RDI: 000000000000000a
[   25.701636][  T228] RBP: 00007ffc3df78510 R08: 0000000000000000 R09: 0000000000300000
[   25.702191][  T228] R10: 0000000000000005 R11: 0000000000000287 R12: 0000000000000000
[   25.702736][  T228] R13: 00007ffc3df78638 R14: 000055a1584aca68 R15: 00007f6d5456a000
[   25.703282][  T228]  </TASK>
[   25.703490][  T228] ==================================================================
[   25.704050][  T228] Disabling lock debugging due to kernel taint

Update copy_from_bpfptr() and strncpy_from_bpfptr() so that:
- for a kernel pointer, it uses the safe copy_from_kernel_nofault() and strncpy_from_kernel_nofault() functions.
- for a userspace pointer, it performs copy_from_user() and strncpy_from_user().

Fixes af2ac3e13e45 ("bpf: Prepare bpf syscall to be used from kernel and user space.")

---
Changelog v2:
Applied Yonghong's comments and directly modified copy_from_bpfptr() and strncpy_from_bpfptr() rather than the sockptr APIs.

 include/linux/bpfptr.h | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
include/linux/bpfptr.h +6 -2
index 46e1757d06a3..79b2f78eec1a 100644
@@ -49,7 +49,9 @@ static inline void bpfptr_add(bpfptr_t *bpfptr, size_t val)
 static inline int copy_from_bpfptr_offset(void *dst, bpfptr_t src,
 					  size_t offset, size_t size)
 {
-	return copy_from_sockptr_offset(dst, (sockptr_t) src, offset, size);
+	if (!bpfptr_is_kernel(src))
+		return copy_from_user(dst, src.user + offset, size);
+	return copy_from_kernel_nofault(dst, src.kernel + offset, size);
 }
 
 static inline int copy_from_bpfptr(void *dst, bpfptr_t src, size_t size)
@@ -78,7 +80,9 @@ static inline void *kvmemdup_bpfptr(bpfptr_t src, size_t len)
 
 static inline long strncpy_from_bpfptr(char *dst, bpfptr_t src, size_t count)
 {
-	return strncpy_from_sockptr(dst, (sockptr_t) src, count);
+	if (bpfptr_is_kernel(src))
+		return strncpy_from_kernel_nofault(dst, src.kernel, count);
+	return strncpy_from_user(dst, src.user, count);
 }
 
 #endif /* _LINUX_BPFPTR_H */

base-commit: 64893e83f916145fd23182209009e9d2ce36d2df

Signature · 1 line
2.35.1
Acked-by on v2 1/1
#1 lore raw
Quoted 82 lines · Jinghao Jia

The bpf_sys_bpf() helper function allows an eBPF program to load another eBPF program from within the kernel. In this case the argument union bpf_attr pointer (as well as the insns and license pointers inside) is a kernel address instead of a userspace address (which is the case of a usual bpf() syscall). To make the memory copying process in the syscall work in both cases, bpfptr_t was introduced to wrap around the pointer and distinguish its origin. Specifically, when copying memory contents from a bpfptr_t, a copy_from_user() is performed in case of a userspace address and a memcpy() is performed for a kernel address.

This can lead to problems because the in-kernel pointer is never checked for validity. The problem happens when an eBPF syscall program tries to call bpf_sys_bpf() to load a program but provides a bad insns pointer -- say 0xdeadbeef -- in the bpf_attr union. The helper calls __sys_bpf() which would then call bpf_prog_load() to load the program. bpf_prog_load() is responsible for copying the eBPF instructions to the newly allocated memory for the program; it creates a kernel bpfptr_t for insns and invokes copy_from_bpfptr(). Internally, all bpfptr_t operations are backed by the corresponding sockptr_t operations, which performs direct memcpy() on kernel pointers for copy_from/strncpy_from operations. Therefore, the code is always happy to dereference the bad pointer to trigger a un-handle-able page fault and in turn an oops. However, this is not supposed to happen because at that point the eBPF program is already verified and should not cause a memory error.

Sample KASAN trace:

[   25.685056][  T228] ==================================================================
[   25.685680][  T228] BUG: KASAN: user-memory-access in copy_from_bpfptr+0x21/0x30
[   25.686210][  T228] Read of size 80 at addr 00000000deadbeef by task poc/228
[   25.686732][  T228]
[   25.686893][  T228] CPU: 3 PID: 228 Comm: poc Not tainted 5.19.0-rc7 #7
[   25.687375][  T228] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS d55cb5a 04/01/2014
[   25.687991][  T228] Call Trace:
[   25.688223][  T228]  <TASK>
[   25.688429][  T228]  dump_stack_lvl+0x73/0x9e
[   25.688747][  T228]  print_report+0xea/0x200
[   25.689061][  T228]  ? copy_from_bpfptr+0x21/0x30
[   25.689401][  T228]  ? _printk+0x54/0x6e
[   25.689693][  T228]  ? _raw_spin_lock_irqsave+0x70/0xd0
[   25.690071][  T228]  ? copy_from_bpfptr+0x21/0x30
[   25.690412][  T228]  kasan_report+0xb5/0xe0
[   25.690716][  T228]  ? copy_from_bpfptr+0x21/0x30
[   25.691059][  T228]  kasan_check_range+0x2bd/0x2e0
[   25.691405][  T228]  ? copy_from_bpfptr+0x21/0x30
[   25.691734][  T228]  memcpy+0x25/0x60
[   25.692000][  T228]  copy_from_bpfptr+0x21/0x30
[   25.692328][  T228]  bpf_prog_load+0x604/0x9e0
[   25.692653][  T228]  ? cap_capable+0xb4/0xe0
[   25.692956][  T228]  ? security_capable+0x4f/0x70
[   25.693324][  T228]  __sys_bpf+0x3af/0x580
[   25.693635][  T228]  bpf_sys_bpf+0x45/0x240
[   25.693937][  T228]  bpf_prog_f0ec79a5a3caca46_bpf_func1+0xa2/0xbd
[   25.694394][  T228]  bpf_prog_run_pin_on_cpu+0x2f/0xb0
[   25.694756][  T228]  bpf_prog_test_run_syscall+0x146/0x1c0
[   25.695144][  T228]  bpf_prog_test_run+0x172/0x190
[   25.695487][  T228]  __sys_bpf+0x2c5/0x580
[   25.695776][  T228]  __x64_sys_bpf+0x3a/0x50
[   25.696084][  T228]  do_syscall_64+0x60/0x90
[   25.696393][  T228]  ? fpregs_assert_state_consistent+0x50/0x60
[   25.696815][  T228]  ? exit_to_user_mode_prepare+0x36/0xa0
[   25.697202][  T228]  ? syscall_exit_to_user_mode+0x20/0x40
[   25.697586][  T228]  ? do_syscall_64+0x6e/0x90
[   25.697899][  T228]  entry_SYSCALL_64_after_hwframe+0x63/0xcd
[   25.698312][  T228] RIP: 0033:0x7f6d543fb759
[   25.698624][  T228] Code: 08 5b 89 e8 5d c3 66 2e 0f 1f 84 00 00 00 00 00 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 97 a6 0e 00 f7 d8 64 89 01 48
[   25.699946][  T228] RSP: 002b:00007ffc3df78468 EFLAGS: 00000287 ORIG_RAX: 0000000000000141
[   25.700526][  T228] RAX: ffffffffffffffda RBX: 00007ffc3df78628 RCX: 00007f6d543fb759
[   25.701071][  T228] RDX: 0000000000000090 RSI: 00007ffc3df78478 RDI: 000000000000000a
[   25.701636][  T228] RBP: 00007ffc3df78510 R08: 0000000000000000 R09: 0000000000300000
[   25.702191][  T228] R10: 0000000000000005 R11: 0000000000000287 R12: 0000000000000000
[   25.702736][  T228] R13: 00007ffc3df78638 R14: 000055a1584aca68 R15: 00007f6d5456a000
[   25.703282][  T228]  </TASK>
[   25.703490][  T228] ==================================================================
[   25.704050][  T228] Disabling lock debugging due to kernel taint

Update copy_from_bpfptr() and strncpy_from_bpfptr() so that:
- for a kernel pointer, it uses the safe copy_from_kernel_nofault() and strncpy_from_kernel_nofault() functions.
- for a userspace pointer, it performs copy_from_user() and strncpy_from_user().

Quoted 3 lines
Fixes af2ac3e13e45 ("bpf: Prepare bpf syscall to be used from kernel and user space.")
#2 lore raw
In reply to Jinghao Jia
This patch was applied to bpf/bpf.git (master) by Alexei Starovoitov <[email protected]>:

Hello:

This patch was applied to bpf/bpf.git (master)
by Alexei Starovoitov <[email protected]>:

Quoted 11 lines · On Fri, 29 Jul 2022 20:17:13 +0000 you wrote:

The bpf_sys_bpf() helper function allows an eBPF program to load another eBPF program from within the kernel. In this case the argument union bpf_attr pointer (as well as the insns and license pointers inside) is a kernel address instead of a userspace address (which is the case of a usual bpf() syscall). To make the memory copying process in the syscall work in both cases, bpfptr_t was introduced to wrap around the pointer and distinguish its origin. Specifically, when copying memory contents from a bpfptr_t, a copy_from_user() is performed in case of a userspace address and a memcpy() is performed for a kernel address.

[...]

Here is the summary with links:
- [v2] BPF: Fix potential bad pointer dereference in bpf_sys_bpf()

    https://git.kernel.org/bpf/bpf/c/e2dcac2f58f5

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html