Skip to contentSkip to conversation

mm/huge_memory: couple fixes and one cleanup

PATCH In review Luis Chamberlain 3 participants 11 messages Apr 24, 2024 – May 7, 2024

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

cover mm/huge_memory: couple fixes and one cleanup

I've been stress testing large folio splits using the new debugfs interface with a new fstests test [0], to make the debugfs interface more useful it helps to do bounds checks on input parameters so this fixes two issues found while doing automatic stress testing of the debugfs interface with found files.

Provided there are no issues I think this should go in for v6.9-rc6.

Also, *should* we strive to support splits for large folios to min order with say MADV_NOHUGEPAGE ?

[0] https://lkml.kernel.org/r/[email protected]

Luis Chamberlain (2):
mm/huge_memory: skip invalid debugfs file entry for folio split mm/huge_memory: cap max length on debugfs file entry folio split

 mm/huge_memory.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)
Signature · 1 line
2.43.0
#1 lore raw

PATCH 1/2 mm/huge_memory: skip invalid debugfs file entry for folio split

If the file entry is too long we may easily end up going out of bounds and crash after strsep() on sscanf(). To avoid this ensure we bound the string to an expected length before we use sscanf() on it.

---
 mm/huge_memory.c | 9 +++++++++
 1 file changed, 9 insertions(+)
mm/huge_memory.c +9 -0
index 9e9879d2f501..8386d24a163e 100644
@@ -3623,6 +3623,7 @@ static ssize_t split_huge_pages_write(struct file *file, const char __user *buf,
 		char file_path[MAX_INPUT_BUF_SZ];
 		pgoff_t off_start = 0, off_end = 0;
 		size_t input_len = strlen(input_buf);
+		size_t max_left_over;
 
 		tok = strsep(&buf, ",");
 		if (tok) {
@@ -3632,6 +3633,14 @@ static ssize_t split_huge_pages_write(struct file *file, const char __user *buf,
 			goto out;
 		}
 
+		max_left_over = MAX_INPUT_BUF_SZ - strlen(file_path);
+		if (!buf ||
+		    strnlen(buf, max_left_over) < 7 ||
+		    strnlen(buf, max_left_over) > max_left_over) {
+			ret = -EINVAL;
+			goto out;
+		}
+
 		ret = sscanf(buf, "0x%lx,0x%lx,%d", &off_start, &off_end, &new_order);
 		if (ret != 2 && ret != 3) {
 			ret = -EINVAL;
Signature · 1 line
2.43.0
reviewer
#2 lore raw

Re: [PATCH 1/2] mm/huge_memory: skip invalid debugfs file entry for folio split

Quoted 22 lines · Luis Chamberlain

If the file entry is too long we may easily end up going out of bounds and crash after strsep() on sscanf(). To avoid this ensure we bound the string to an expected length before we use sscanf() on it.

---
 mm/huge_memory.c | 9 +++++++++
 1 file changed, 9 insertions(+)
mm/huge_memory.c +1 -0
index 9e9879d2f501..8386d24a163e 100644
@@ -3623,6 +3623,7 @@ static ssize_t split_huge_pages_write(struct file *file, const char __user *buf,
 		char file_path[MAX_INPUT_BUF_SZ];
 		pgoff_t off_start = 0, off_end = 0;
 		size_t input_len = strlen(input_buf);
+		size_t max_left_over;
 
 		tok = strsep(&buf, ",");
 		if (tok) {

mm/huge_memory.c static ssize_t split_huge_pages_write(struct file *file, const char __user *buf, · line 3638

Quoted 6 lines
mm/huge_memory.c +3 -0
@@ -3632,6 +3633,14 @@ static ssize_t split_huge_pages_write(struct file *file, const char __user *buf,
 			goto out;
 		}
 
+		max_left_over = MAX_INPUT_BUF_SZ - strlen(file_path);
+		if (!buf ||
+		    strnlen(buf, max_left_over) < 7 ||

What is this magic number 7? strlen("0xN,0xN") as the minimal input string size? Maybe use sizeof("0xN,0xN") - 1 instead?

Quoted 10 lines
mm/huge_memory.c +5 -0
+		    strnlen(buf, max_left_over) > max_left_over) {
+			ret = -EINVAL;
+			goto out;
+		}
+
 		ret = sscanf(buf, "0x%lx,0x%lx,%d", &off_start, &off_end, &new_order);
 		if (ret != 2 && ret != 3) {
 			ret = -EINVAL;
Signature · 1 line
2.43.0

Everything else looks good to me. Reviewed-by: Zi Yan <[email protected]>

--
Best Regards,
Yan, Zi

#3 lore raw

Re: [PATCH 1/2] mm/huge_memory: skip invalid debugfs file entry for folio split

Quoted 27 lines · Zi Yan
Quoted 25 lines · Luis Chamberlain

If the file entry is too long we may easily end up going out of bounds and crash after strsep() on sscanf(). To avoid this ensure we bound the string to an expected length before we use sscanf() on it.

---
 mm/huge_memory.c | 9 +++++++++
 1 file changed, 9 insertions(+)
mm/huge_memory.c +1 -0
index 9e9879d2f501..8386d24a163e 100644
@@ -3623,6 +3623,7 @@ static ssize_t split_huge_pages_write(struct file *file, const char __user *buf,
 		char file_path[MAX_INPUT_BUF_SZ];
 		pgoff_t off_start = 0, off_end = 0;
 		size_t input_len = strlen(input_buf);
+		size_t max_left_over;
 
 		tok = strsep(&buf, ",");
 		if (tok) {
@@ -3632,6 +3633,14 @@ static ssize_t split_huge_pages_write(struct file *file, const char __user *buf,
 			goto out;
 		}
 
Quoted 6 lines

mm/huge_memory.c static ssize_t split_huge_pages_write(struct file *file, const char __user *buf, · line 3638

Quoted 3 lines
mm/huge_memory.c +3 -0
+		max_left_over = MAX_INPUT_BUF_SZ - strlen(file_path);
+		if (!buf ||
+		    strnlen(buf, max_left_over) < 7 ||

What is this magic number 7? strlen("0xN,0xN") as the minimal input string size? Maybe use sizeof("0xN,0xN") - 1 instead?

Sure and I forgot the fixes tag, will send a v2.

Luis

maintainer
#4 lore raw

Re: [PATCH 1/2] mm/huge_memory: skip invalid debugfs file entry for folio split

In reply to Luis Chamberlain
Quoted 3 lines · On Wed, 24 Apr 2024 15:54:48 -0700 Luis Chamberlain <[email protected]> wrote:

If the file entry is too long we may easily end up going out of bounds and crash after strsep() on sscanf().

Can you explain why? I'm not seeing it.

#5 lore raw

Re: [PATCH 1/2] mm/huge_memory: skip invalid debugfs file entry for folio split

Quoted 7 lines · Andrew Morton
Quoted 3 lines · On Wed, 24 Apr 2024 15:54:48 -0700 Luis Chamberlain <[email protected]> wrote:

If the file entry is too long we may easily end up going out of bounds and crash after strsep() on sscanf().

Can you explain why? I'm not seeing it.

I couldn't see it either but I just looked at the crash below and its the only thing I could think of. So I think its when userspace somehow abuses MAX_INPUT_BUF_SZ a lot somehow.

The test I am using:

https://lore.kernel.org/all/[email protected]/

The crash:

Apr 28 20:09:21 debian12-xfs-reflink-16k-4ks unknown: run fstests generic/745 at 2024-04-28 20:09:21
Apr 28 20:09:22 debian12-xfs-reflink-16k-4ks kernel: XFS (loop5): EXPERIMENTAL: Filesystem with Large Block Size (16384 bytes) enabled.
Apr 28 20:09:22 debian12-xfs-reflink-16k-4ks kernel: XFS (loop5): Mounting V5 Filesystem 6633551d-bd66-4a7c-a4ce-7e312e611742
Apr 28 20:09:22 debian12-xfs-reflink-16k-4ks kernel: XFS (loop5): Ending clean mount
Apr 28 20:09:25 debian12-xfs-reflink-16k-4ks kernel: huge_memory: split file-backed THPs in file: /media/scratch/test/fss896, page offset: [0x0 - 0x0] with order: 2
Apr 28 20:09:25 debian12-xfs-reflink-16k-4ks kernel: huge_memory: 0 of 0 file-backed THP split
Apr 28 20:09:25 debian12-xfs-reflink-16k-4ks kernel: huge_memory: split file-backed THPs in file: /media/scratch/test/p0/f0, page offset: [0x0 - 0x6a4000] with order: 2
Apr 28 20:09:25 debian12-xfs-reflink-16k-4ks kernel: huge_memory: 0 of 2 file-backed THP split
Apr 28 20:09:25 debian12-xfs-reflink-16k-4ks kernel: huge_memory: split file-backed THPs in file: /media/scratch/test/p0/f1, page offset: [0x0 - 0x0] with order: 2
Apr 28 20:09:25 debian12-xfs-reflink-16k-4ks kernel: huge_memory: 0 of 0 file-backed THP split
Apr 28 20:09:25 debian12-xfs-reflink-16k-4ks kernel: huge_memory: split file-backed THPs in file: /media/scratch/test/p2/d0/d1/d4/f5, page offset: [0x0 - 0xc83d64] with order: 2
Apr 28 20:09:25 debian12-xfs-reflink-16k-4ks kernel: huge_memory: 1 of 6 file-backed THP split
Apr 28 20:09:25 debian12-xfs-reflink-16k-4ks kernel: huge_memory: split file-backed THPs in file: /media/scratch/test/p2/d0/d1/d4/f9, page offset: [0x0 - 0xcad301] with order: 2
Apr 28 20:09:25 debian12-xfs-reflink-16k-4ks kernel: huge_memory: 9 of 79 file-backed THP split

<--- after ~2 minutes -->

Apr 28 20:11:24 debian12-xfs-reflink-16k-4ks kernel: huge_memory: split file-backed THPs in file: /media/scratch/test/p0/dc/dd/d274/d2ca/de12/d139/d4b1/d3cf/d4bb/d1240/d96c/d902/d7a5/d8b0/dbfb/dd18/d795/d8a1/dcc5/d1950/d19ed/d1204/d1452/db5f/dbb7/d10e8/d316/d707/f8f3, page offset: [0x0 - 0x17c0000] with order: 2
Apr 28 20:11:24 debian12-xfs-reflink-16k-4ks kernel: huge_memory: 0 of 6 file-backed THP split
Apr 28 20:11:24 debian12-xfs-reflink-16k-4ks kernel: huge_memory: split file-backed THPs in file: /media/scratch/test/p0/dc/dd/d274/d2ca/de12/d139/d4b1/d3cf/d4bb/d1240/d96c/d902/d7a5/d8b0/dbfb/dd18/d795/d8a1/dcc5/d1950/d19ed/d1204/d1452/db5f/dbb7/d10e8/d316/d707/fbb8, page offset: [0x0 - 0x32b533] with order: 2
Apr 28 20:11:24 debian12-xfs-reflink-16k-4ks kernel: huge_memory: 0 of 5 file-backed THP split
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: BUG: kernel NULL pointer dereference, address: 0000000000000000
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: #PF: supervisor read access in kernel mode
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: #PF: error_code(0x0000) - not-present page
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: PGD 0 P4D 0
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: Oops: 0000 [#1] PREEMPT SMP NOPTI
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: CPU: 7 PID: 2177 Comm: 745 Not tainted 6.9.0-rc5+ #5
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: RIP: 0010:vsscanf (lib/vsprintf.c:3465) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: Code: 0f 85 1d 02 00 00 48 8b 54 24 08 0f b6 02 3c 25 40 0f 95 c7 84 c0 0f 95 c1 40 20 cf 41 89 fe 74 40 48 8d 4a 01 48 89 4c 24 08 <41> 0f b6 45 00 38 02 0f 84 87 02 00 00 48 8b 44 24 38 65 48 2b 04
All code
========
   0:	0f 85 1d 02 00 00    	jne    0x223
   6:	48 8b 54 24 08       	mov    0x8(%rsp),%rdx
   b:	0f b6 02             	movzbl (%rdx),%eax
   e:	3c 25                	cmp    $0x25,%al
  10:	40 0f 95 c7          	setne  %dil
  14:	84 c0                	test   %al,%al
  16:	0f 95 c1             	setne  %cl
  19:	40 20 cf             	and    %cl,%dil
  1c:	41 89 fe             	mov    %edi,%r14d
  1f:	74 40                	je     0x61
  21:	48 8d 4a 01          	lea    0x1(%rdx),%rcx
  25:	48 89 4c 24 08       	mov    %rcx,0x8(%rsp)
  2a:*	41 0f b6 45 00       	movzbl 0x0(%r13),%eax		<-- trapping instruction
  2f:	38 02                	cmp    %al,(%rdx)
  31:	0f 84 87 02 00 00    	je     0x2be
  37:	48 8b 44 24 38       	mov    0x38(%rsp),%rax
  3c:	65                   	gs
  3d:	48                   	rex.W
  3e:	2b                   	.byte 0x2b
  3f:	04                   	.byte 0x4

Code starting with the faulting instruction
===========================================

   0:	41 0f b6 45 00       	movzbl 0x0(%r13),%eax
   5:	38 02                	cmp    %al,(%rdx)
   7:	0f 84 87 02 00 00    	je     0x294
   d:	48 8b 44 24 38       	mov    0x38(%rsp),%rax
  12:	65                   	gs
  13:	48                   	rex.W
  14:	2b                   	.byte 0x2b
  15:	04                   	.byte 0x4
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: RSP: 0018:ffffb24ec358bad8 EFLAGS: 00010202
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: RAX: 0000000000000030 RBX: ffffb24ec358bb50 RCX: ffffffffa03fbc18
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: RDX: ffffffffa03fbc17 RSI: ffffffffa03fbc17 RDI: 0000000000000001
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: RBP: 0000000000000000 R08: ffffb24ec358bbdc R09: 000000000000002c
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: R10: ffffb24ec358bbfa R11: 0000000000000000 R12: 0000000000000000
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: R13: 0000000000000000 R14: 0000000000000001 R15: ffffb24ec358bf08
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: FS:  00007f8b39189740(0000) GS:ffff88837bdc0000(0000) knlGS:0000000000000000
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: CR2: 0000000000000000 CR3: 000000010c9f2003 CR4: 0000000000770ef0
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: DR3: 0000000000000000 DR6: 00000000fffe07f0 DR7: 0000000000000400
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: PKRU: 55555554
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: Call Trace:
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel:  <TASK>
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? __die (arch/x86/kernel/dumpstack.c:421 arch/x86/kernel/dumpstack.c:434) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? page_fault_oops (arch/x86/mm/fault.c:713) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? do_user_addr_fault (./include/linux/kprobes.h:591 (discriminator 1) arch/x86/mm/fault.c:1265 (discriminator 1)) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? exc_page_fault (./arch/x86/include/asm/paravirt.h:693 arch/x86/mm/fault.c:1513 arch/x86/mm/fault.c:1563) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? asm_exc_page_fault (./arch/x86/include/asm/idtentry.h:623) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? vsscanf (lib/vsprintf.c:3465) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? stack_depot_save_flags (lib/stackdepot.c:609) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: sscanf (lib/vsprintf.c:3725) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? __check_object_size (mm/usercopy.c:226 (discriminator 1) mm/usercopy.c:213 (discriminator 1)) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: split_huge_pages_write (mm/huge_memory.c:3664) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: full_proxy_write (fs/debugfs/file.c:328 (discriminator 1)) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: vfs_write (fs/read_write.c:588) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? __count_memcg_events (mm/memcontrol.c:723 (discriminator 1) mm/memcontrol.c:962 (discriminator 1)) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? count_memcg_events.constprop.0 (./arch/x86/include/asm/irqflags.h:134 (discriminator 1) ./include/linux/memcontrol.h:1097 (discriminator 1)) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? handle_mm_fault (mm/memory.c:5483 mm/memory.c:5622) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ? preempt_count_add (./include/linux/ftrace.h:974 (discriminator 1) kernel/sched/core.c:5852 (discriminator 1) kernel/sched/core.c:5849 (discriminator 1) kernel/sched/core.c:5877 (discriminator 1)) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: ksys_write (fs/read_write.c:643) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: do_syscall_64 (arch/x86/entry/common.c:52 (discriminator 1) arch/x86/entry/common.c:83 (discriminator 1)) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:130) 
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: RIP: 0033:0x7f8b39284240
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: Code: 40 00 48 8b 15 c1 9b 0d 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b7 0f 1f 00 80 3d a1 23 0e 00 00 74 17 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 58 c3 0f 1f 80 00 00 00 00 48 83 ec 28 48 89
All code
========
   0:	40 00 48 8b          	rex add %cl,-0x75(%rax)
   4:	15 c1 9b 0d 00       	adc    $0xd9bc1,%eax
   9:	f7 d8                	neg    %eax
   b:	64 89 02             	mov    %eax,%fs:(%rdx)
   e:	48 c7 c0 ff ff ff ff 	mov    $0xffffffffffffffff,%rax
  15:	eb b7                	jmp    0xffffffffffffffce
  17:	0f 1f 00             	nopl   (%rax)
  1a:	80 3d a1 23 0e 00 00 	cmpb   $0x0,0xe23a1(%rip)        # 0xe23c2
  21:	74 17                	je     0x3a
  23:	b8 01 00 00 00       	mov    $0x1,%eax
  28:	0f 05                	syscall
  2a:*	48 3d 00 f0 ff ff    	cmp    $0xfffffffffffff000,%rax		<-- trapping instruction
  30:	77 58                	ja     0x8a
  32:	c3                   	ret
  33:	0f 1f 80 00 00 00 00 	nopl   0x0(%rax)
  3a:	48 83 ec 28          	sub    $0x28,%rsp
  3e:	48                   	rex.W
  3f:	89                   	.byte 0x89

Code starting with the faulting instruction
===========================================

   0:	48 3d 00 f0 ff ff    	cmp    $0xfffffffffffff000,%rax
   6:	77 58                	ja     0x60
   8:	c3                   	ret
   9:	0f 1f 80 00 00 00 00 	nopl   0x0(%rax)
  10:	48 83 ec 28          	sub    $0x28,%rsp
  14:	48                   	rex.W
  15:	89                   	.byte 0x89
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: RSP: 002b:00007fff2a14b5f8 EFLAGS: 00000202 ORIG_RAX: 0000000000000001
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: RAX: ffffffffffffffda RBX: 0000000000000109 RCX: 00007f8b39284240
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: RDX: 0000000000000109 RSI: 00005589ea673400 RDI: 0000000000000001
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: RBP: 00005589ea673400 R08: 0000000000000007 R09: 0000000000000073
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: R10: 0000000000000000 R11: 0000000000000202 R12: 0000000000000109
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel: R13: 00007f8b3935f760 R14: 0000000000000109 R15: 00007f8b3935a9e0
Apr 28 20:11:25 debian12-xfs-reflink-16k-4ks kernel:  </TASK>
maintainer
#6 lore raw

Re: [PATCH 1/2] mm/huge_memory: skip invalid debugfs file entry for folio split

Quoted 7 lines · On Sun, 28 Apr 2024 21:04:50 -0700 Luis Chamberlain <[email protected]> wrote:
Quoted 6 lines · Andrew Morton
Quoted 3 lines · On Wed, 24 Apr 2024 15:54:48 -0700 Luis Chamberlain <[email protected]> wrote:

If the file entry is too long we may easily end up going out of bounds and crash after strsep() on sscanf().

Quoted 5 lines
Quoted 1 line

Can you explain why? I'm not seeing it.

I couldn't see it either but I just looked at the crash below and its the only thing I could think of. So I think its when userspace somehow abuses MAX_INPUT_BUF_SZ a lot somehow.

This isn't a good basis for making kernel changes :(

Can you investigate a little further please? What actually is present at *buf when your new checks succeed? Could we be seeing 0xNNN,0xNNN and leaving new_order unaltered? Or something.

#7 lore raw

Re: [PATCH 1/2] mm/huge_memory: skip invalid debugfs file entry for folio split

Quoted 11 lines · Andrew Morton
Quoted 9 lines · On Sun, 28 Apr 2024 21:04:50 -0700 Luis Chamberlain <[email protected]> wrote:
Quoted 7 lines · Andrew Morton
Quoted 3 lines · On Wed, 24 Apr 2024 15:54:48 -0700 Luis Chamberlain <[email protected]> wrote:

If the file entry is too long we may easily end up going out of bounds and crash after strsep() on sscanf().

Can you explain why? I'm not seeing it.

Quoted 7 lines
Quoted 3 lines

I couldn't see it either but I just looked at the crash below and its the only thing I could think of. So I think its when userspace somehow abuses MAX_INPUT_BUF_SZ a lot somehow.

This isn't a good basis for making kernel changes :(

Can you investigate a little further please?

Sure, this will require some more time, feel free to ignore these two patches for now then.

Luis

#8 lore raw

PATCH 2/2 mm/huge_memory: cap max length on debugfs file entry folio split

In reply to Luis Chamberlain

Don't allow to query beyond a mapped file's length. Since this is just a debugfs interface allow userspace to be lazy and use a large value so we can just use the entire file.

Without this we can end up wasting cycles looking for folios which just don't exist for no good reason.

---
 mm/huge_memory.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
mm/huge_memory.c +6 -3
index 8386d24a163e..86a8c7b3b8dc 100644
@@ -3535,7 +3535,7 @@ static int split_huge_pages_in_file(const char *file_path, pgoff_t off_start,
 	struct file *candidate;
 	struct address_space *mapping;
 	int ret = -EINVAL;
-	pgoff_t index;
+	pgoff_t index, fsize;
 	int nr_pages = 1;
 	unsigned long total = 0, split = 0;
 
@@ -3547,11 +3547,14 @@ static int split_huge_pages_in_file(const char *file_path, pgoff_t off_start,
 	if (IS_ERR(candidate))
 		goto out;
 
+	mapping = candidate->f_mapping;
+	fsize = i_size_read(mapping->host);
+	if (off_end > fsize)
+		off_end = fsize;
+
 	pr_debug("split file-backed THPs in file: %s, page offset: [0x%lx - 0x%lx]\n",
 		 file_path, off_start, off_end);
 
-	mapping = candidate->f_mapping;
-
 	for (index = off_start; index < off_end; index += nr_pages) {
 		struct folio *folio = filemap_get_folio(mapping, index);
 
Signature · 1 line
2.43.0
reviewer
#9 lore raw

Re: [PATCH 2/2] mm/huge_memory: cap max length on debugfs file entry folio split

Quoted 31 lines · Luis Chamberlain

Don't allow to query beyond a mapped file's length. Since this is just a debugfs interface allow userspace to be lazy and use a large value so we can just use the entire file.

Without this we can end up wasting cycles looking for folios which just don't exist for no good reason.

---
 mm/huge_memory.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
mm/huge_memory.c +3 -1
index 8386d24a163e..86a8c7b3b8dc 100644
@@ -3535,7 +3535,7 @@ static int split_huge_pages_in_file(const char *file_path, pgoff_t off_start,
 	struct file *candidate;
 	struct address_space *mapping;
 	int ret = -EINVAL;
-	pgoff_t index;
+	pgoff_t index, fsize;
 	int nr_pages = 1;
 	unsigned long total = 0, split = 0;
 
@@ -3547,11 +3547,14 @@ static int split_huge_pages_in_file(const char *file_path, pgoff_t off_start,
 	if (IS_ERR(candidate))
 		goto out;
 
+	mapping = candidate->f_mapping;
+	fsize = i_size_read(mapping->host);

mm/huge_memory.c static int split_huge_pages_in_file(const char *file_path, pgoff_t off_start, · line 3559

Quoted 10 lines
mm/huge_memory.c +3 -2
+	if (off_end > fsize)
+		off_end = fsize;
+
 	pr_debug("split file-backed THPs in file: %s, page offset: [0x%lx - 0x%lx]\n",
 		 file_path, off_start, off_end);
 
-	mapping = candidate->f_mapping;
-
 	for (index = off_start; index < off_end; index += nr_pages) {
 		struct folio *folio = filemap_get_folio(mapping, index);

LGTM. Reviewed-by: Zi Yan <[email protected]>

--
Best Regards,
Yan, Zi