Re: [Patch bpf] vmalloc: relax is_vmalloc_or_module_addr() check

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 5/27/24 3:28 PM, Mike Rapoport wrote:
On Mon, May 27, 2024 at 03:11:50PM +0200, Daniel Borkmann wrote:
On 5/27/24 1:06 AM, Cong Wang wrote:
From: Cong Wang <cong.wang@xxxxxxxxxxxxx>

After commit 2c9e5d4a0082 ("bpf: remove CONFIG_BPF_JIT dependency on CONFIG_MODULES of")
CONFIG_BPF_JIT does not depend on CONFIG_MODULES any more and bpf jit
also uses the MODULES_VADDR ~ MODULES_END memory region. But
is_vmalloc_or_module_addr() still checks CONFIG_MODULES, which then
returns false for a bpf jit memory region when CONFIG_MODULES is not
defined. It leads to the following kernel BUG:

[    1.567023] ------------[ cut here ]------------
[    1.567883] kernel BUG at mm/vmalloc.c:745!
[    1.568477] Oops: invalid opcode: 0000 [#1] PREEMPT SMP KASAN NOPTI
[    1.569367] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 6.9.0+ #448
[    1.570247] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.15.0-1 04/01/2014
[    1.570786] RIP: 0010:vmalloc_to_page+0x48/0x1ec
[    1.570786] Code: 0f 00 00 e8 eb 1a 05 00 b8 37 00 00 00 48 ba fe ff ff ff ff 1f 00 00 4c 03 25 76 49 c6 02 48 c1 e0 28 48 01 e8 48 39 d0 76 02 <0f> 0b 4c 89 e7 e8 bf 1a 05 00 49 8b 04 24 48 a9 9f ff ff ff 0f 84
[    1.570786] RSP: 0018:ffff888007787960 EFLAGS: 00010212
[    1.570786] RAX: 000036ffa0000000 RBX: 0000000000000640 RCX: ffffffff8147e93c
[    1.570786] RDX: 00001ffffffffffe RSI: dffffc0000000000 RDI: ffffffff840e32c8
[    1.570786] RBP: ffffffffa0000000 R08: 0000000000000000 R09: 0000000000000000
[    1.570786] R10: ffff888007787a88 R11: ffffffff8475d8e7 R12: ffffffff83e80ff8
[    1.570786] R13: 0000000000000640 R14: 0000000000000640 R15: 0000000000000640
[    1.570786] FS:  0000000000000000(0000) GS:ffff88806cc00000(0000) knlGS:0000000000000000
[    1.570786] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    1.570786] CR2: ffff888006a01000 CR3: 0000000003e80000 CR4: 0000000000350ef0
[    1.570786] Call Trace:
[    1.570786]  <TASK>
[    1.570786]  ? __die_body+0x1b/0x58
[    1.570786]  ? die+0x31/0x4b
[    1.570786]  ? do_trap+0x9d/0x138
[    1.570786]  ? vmalloc_to_page+0x48/0x1ec
[    1.570786]  ? do_error_trap+0xcd/0x102
[    1.570786]  ? vmalloc_to_page+0x48/0x1ec
[    1.570786]  ? vmalloc_to_page+0x48/0x1ec
[    1.570786]  ? handle_invalid_op+0x2f/0x38
[    1.570786]  ? vmalloc_to_page+0x48/0x1ec
[    1.570786]  ? exc_invalid_op+0x2b/0x41
[    1.570786]  ? asm_exc_invalid_op+0x16/0x20
[    1.570786]  ? vmalloc_to_page+0x26/0x1ec
[    1.570786]  ? vmalloc_to_page+0x48/0x1ec
[    1.570786]  __text_poke+0xb6/0x458
[    1.570786]  ? __pfx_text_poke_memcpy+0x10/0x10
[    1.570786]  ? __pfx___mutex_lock+0x10/0x10
[    1.570786]  ? __pfx___text_poke+0x10/0x10
[    1.570786]  ? __pfx_get_random_u32+0x10/0x10
[    1.570786]  ? srso_return_thunk+0x5/0x5f
[    1.570786]  text_poke_copy_locked+0x70/0x84
[    1.570786]  text_poke_copy+0x32/0x4f
[    1.570786]  bpf_arch_text_copy+0xf/0x27
[    1.570786]  bpf_jit_binary_pack_finalize+0x26/0x5a
[    1.570786]  bpf_int_jit_compile+0x576/0x8ad
[    1.570786]  ? __pfx_bpf_int_jit_compile+0x10/0x10
[    1.570786]  ? srso_return_thunk+0x5/0x5f
[    1.570786]  ? __kmalloc_node_track_caller+0x2b5/0x2e0
[    1.570786]  bpf_prog_select_runtime+0x7c/0x199
[    1.570786]  bpf_prepare_filter+0x1e9/0x25b
[    1.570786]  ? __pfx_bpf_prepare_filter+0x10/0x10
[    1.570786]  ? srso_return_thunk+0x5/0x5f
[    1.570786]  ? _find_next_bit+0x29/0x7e
[    1.570786]  bpf_prog_create+0xb8/0xe0
[    1.570786]  ptp_classifier_init+0x75/0xa1
[    1.570786]  ? __pfx_ptp_classifier_init+0x10/0x10
[    1.570786]  ? srso_return_thunk+0x5/0x5f
[    1.570786]  ? register_pernet_subsys+0x36/0x42
[    1.570786]  ? srso_return_thunk+0x5/0x5f
[    1.570786]  sock_init+0x99/0xa3
[    1.570786]  ? __pfx_sock_init+0x10/0x10
[    1.570786]  do_one_initcall+0x104/0x2c4
[    1.570786]  ? __pfx_do_one_initcall+0x10/0x10
[    1.570786]  ? parameq+0x25/0x2d
[    1.570786]  ? rcu_is_watching+0x1c/0x3c
[    1.570786]  ? trace_kmalloc+0x81/0xb2
[    1.570786]  ? srso_return_thunk+0x5/0x5f
[    1.570786]  ? __kmalloc+0x29c/0x2c7
[    1.570786]  ? srso_return_thunk+0x5/0x5f
[    1.570786]  do_initcalls+0xf9/0x123
[    1.570786]  kernel_init_freeable+0x24f/0x289
[    1.570786]  ? __pfx_kernel_init+0x10/0x10
[    1.570786]  kernel_init+0x19/0x13a
[    1.570786]  ret_from_fork+0x24/0x41
[    1.570786]  ? __pfx_kernel_init+0x10/0x10
[    1.570786]  ret_from_fork_asm+0x1a/0x30
[    1.570786]  </TASK>
[    1.570819] ---[ end trace 0000000000000000 ]---
[    1.571463] RIP: 0010:vmalloc_to_page+0x48/0x1ec
[    1.572111] Code: 0f 00 00 e8 eb 1a 05 00 b8 37 00 00 00 48 ba fe ff ff ff ff 1f 00 00 4c 03 25 76 49 c6 02 48 c1 e0 28 48 01 e8 48 39 d0 76 02 <0f> 0b 4c 89 e7 e8 bf 1a 05 00 49 8b 04 24 48 a9 9f ff ff ff 0f 84
[    1.574632] RSP: 0018:ffff888007787960 EFLAGS: 00010212
[    1.575129] RAX: 000036ffa0000000 RBX: 0000000000000640 RCX: ffffffff8147e93c
[    1.576097] RDX: 00001ffffffffffe RSI: dffffc0000000000 RDI: ffffffff840e32c8
[    1.577084] RBP: ffffffffa0000000 R08: 0000000000000000 R09: 0000000000000000
[    1.578077] R10: ffff888007787a88 R11: ffffffff8475d8e7 R12: ffffffff83e80ff8
[    1.578810] R13: 0000000000000640 R14: 0000000000000640 R15: 0000000000000640
[    1.579823] FS:  0000000000000000(0000) GS:ffff88806cc00000(0000) knlGS:0000000000000000
[    1.580992] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    1.581869] CR2: ffff888006a01000 CR3: 0000000003e80000 CR4: 0000000000350ef0
[    1.582800] Kernel panic - not syncing: Fatal exception
[    1.583765] ---[ end Kernel panic - not syncing: Fatal exception ]---

Fixes: 2c9e5d4a0082 ("bpf: remove CONFIG_BPF_JIT dependency on CONFIG_MODULES of")
Cc: Luis Chamberlain <mcgrof@xxxxxxxxxx>
Cc: Mike Rapoport (IBM) <rppt@xxxxxxxxxx>
Signed-off-by: Cong Wang <cong.wang@xxxxxxxxxxxxx>

Thanks for the fix!

Mike/Luis, do you plan to pick this up or rather prefer if we route it to
Linus (with your Ack assuming it looks good to you)?

I'm fine with routing this via bpf, but usually vmalloc patches go via mm
tree, so it's more up to Andrew.

All good, so once Cong sends v2, lets have Andrew pick up the fix.

Thanks,
Daniel




[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux