* [BUG] Warning in ext4_check_map_extents_env
@ 2026-03-09 13:41 Pedro Demarchi Gomes
2026-03-10 12:23 ` Zhang Yi
0 siblings, 1 reply; 3+ messages in thread
From: Pedro Demarchi Gomes @ 2026-03-09 13:41 UTC (permalink / raw)
To: Theodore Ts'o, Andreas Dilger; +Cc: linux-ext4
Hi,
The warning below occurs when running selftests/mm/merge on a kernel built
with CONFIG_EXT4_DEBUG=y.
syzkaller login: [ 29.277674] ------------[ cut here ]------------
[ 29.279636] WARNING: fs/ext4/inode.c:436 at ext4_check_map_extents_env+0x389/0x440, CPU#1: merge/365
[ 29.282459] Modules linked in:
[ 29.283497] CPU: 1 UID: 0 PID: 365 Comm: merge Not tainted 7.0.0-rc3 #59 PREEMPT(full)
[ 29.285999] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[ 29.288850] RIP: 0010:ext4_check_map_extents_env+0x389/0x440
[ 29.290653] Code: ff 48 89 da 48 b8 00 00 00 00 00 fc ff df 48 c1 ea 03 80 3c 02 00 0f 85 87 00 00 00 48 8b 45 18 48 85 c0 0f 85 2e fd ff ff 90 <0f> 0b 90 e9 25 fd ff ff e8 8a 2e ca ff e9 e4 fc ff ff 48 89 df e8
[ 29.296623] RSP: 0018:ffff88811876f2a8 EFLAGS: 00010246
[ 29.298288] RAX: 0000000000000000 RBX: ffff888106c66798 RCX: ffffffff81e8ad60
[ 29.300529] RDX: 1ffff11020d8ccf3 RSI: 0000000000000008 RDI: ffff888106c66798
[ 29.302748] RBP: ffff888106c66780 R08: 0000000000000000 R09: ffffed1020d8ccf3
[ 29.305009] R10: ffff888106c6679f R11: 0000000000000001 R12: 00000000001e0c26
[ 29.307226] R13: ffff88811876f494 R14: ffff88811876f498 R15: ffff88811876f488
[ 29.309477] FS: 00007fbe82723780(0000) GS:ffff8883c38aa000(0000) knlGS:0000000000000000
[ 29.312085] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 29.313749] CR2: 00007fbe82790915 CR3: 000000011880e000 CR4: 00000000000006f0
[ 29.315643] Call Trace:
[ 29.316170] <TASK>
[ 29.316539] ext4_map_blocks+0x602/0xb40
[ 29.317204] ? __pfx_ext4_map_blocks+0x10/0x10
[ 29.317971] ? __pfx_stack_trace_save+0x10/0x10
[ 29.318719] ? __filemap_add_folio+0x66f/0xb30
[ 29.319500] ext4_mpage_readpages.constprop.0.isra.0+0x918/0xe20
[ 29.320501] ? __pfx_ext4_mpage_readpages.constprop.0.isra.0+0x10/0x10
[ 29.321602] ? filemap_add_folio+0x11e/0x2c0
[ 29.322325] ? __pfx_ext4_read_folio+0x10/0x10
[ 29.323107] ext4_read_folio+0xd9/0x230
[ 29.323762] ? __pfx_ext4_read_folio+0x10/0x10
[ 29.324552] ? __pfx_ext4_read_folio+0x10/0x10
[ 29.325321] filemap_read_folio+0x43/0x150
[ 29.326041] do_read_cache_folio+0x1b1/0x320
[ 29.326759] read_cache_page+0x46/0x120
[ 29.327426] install_breakpoint+0x27d/0x810
[ 29.328123] uprobe_mmap+0x439/0xeb0
[ 29.328715] ? __pfx_uprobe_mmap+0x10/0x10
[ 29.329391] __mmap_region+0x8af/0x2290
[ 29.330074] ? __pfx___mmap_region+0x10/0x10
[ 29.330793] ? unwind_get_return_address+0x59/0xa0
[ 29.331630] ? event_sched_in+0x33f/0xaf0
[ 29.332308] ? perf_event_groups_next+0x9a/0x200
[ 29.333667] ? mm_get_unmapped_area_vmflags+0x6c/0xe0
[ 29.335295] ? mmap_region+0xd3/0x2b0
[ 29.336171] do_mmap+0x98a/0xec0
[ 29.336713] ? ipe_mmap_file+0xcf/0xe0
[ 29.337340] ? __pfx_do_mmap+0x10/0x10
[ 29.337977] ? __pfx_down_write_killable+0x10/0x10
[ 29.338757] ? __pfx_mutex_unlock+0x10/0x10
[ 29.339477] vm_mmap_pgoff+0x1b2/0x310
[ 29.340114] ? __pfx_vm_mmap_pgoff+0x10/0x10
[ 29.340819] ? __pfx___do_sys_perf_event_open+0x10/0x10
[ 29.341688] ? fget+0x189/0x230
[ 29.342223] ksys_mmap_pgoff+0x2b1/0x5b0
[ 29.342868] ? __pfx_ksys_mmap_pgoff+0x10/0x10
[ 29.343618] do_syscall_64+0xe2/0x560
[ 29.344246] entry_SYSCALL_64_after_hwframe+0x77/0x7f
[ 29.345105] RIP: 0033:0x7fbe82832de2
[ 29.345707] Code: 00 00 00 0f 1f 44 00 00 41 f7 c1 ff 0f 00 00 75 27 55 89 cd 53 48 89 fb 48 85 ff 74 3b 41 89 ea 48 89 df b8 09 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 76 5b 5d c3 0f 1f 00 48 8b 05 f9 8f 0d 00 64
[ 29.348851] RSP: 002b:00007ffd535f3b38 EFLAGS: 00000206 ORIG_RAX: 0000000000000009
[ 29.350309] RAX: ffffffffffffffda RBX: 00007fbe826f9000 RCX: 00007fbe82832de2
[ 29.351526] RDX: 0000000000000004 RSI: 000000000000a000 RDI: 00007fbe826f9000
[ 29.352865] RBP: 0000000000000012 R08: 000000000000000b R09: 0000000000000000
[ 29.354136] R10: 0000000000000012 R11: 0000000000000206 R12: 00007ffd535f4120
[ 29.355357] R13: 0000000000000000 R14: 000000000000000b R15: 0000000000001000
[ 29.356684] </TASK>
[ 29.357119] ---[ end trace 0000000000000000 ]---
The warning is triggered by TEST_F(merge, handle_uprobe_upon_merged_vma) in
selftests/mm/merge.c. It occurs during the first mmap() following the
__NR_perf_event_open syscall, as shown in the excerpt below:
TEST_F(merge, handle_uprobe_upon_merged_vma)
{
const size_t attr_sz = sizeof(struct perf_event_attr);
unsigned int page_size = self->page_size;
const char *probe_file = "./foo";
char *carveout = self->carveout;
struct perf_event_attr attr;
unsigned long type;
void *ptr1, *ptr2;
int fd;
fd = open(probe_file, O_RDWR|O_CREAT, 0600);
ASSERT_GE(fd, 0);
ASSERT_EQ(ftruncate(fd, page_size), 0);
if (read_sysfs("/sys/bus/event_source/devices/uprobe/type", &type) != 0) {
SKIP(goto out, "Failed to read uprobe sysfs file, skipping");
}
memset(&attr, 0, attr_sz);
attr.size = attr_sz;
attr.type = type;
attr.config1 = (__u64)(long)probe_file;
attr.config2 = 0x0;
ASSERT_GE(syscall(__NR_perf_event_open, &attr, 0, -1, -1, 0), 0);
ptr1 = mmap(&carveout[page_size], 10 * page_size, PROT_EXEC,
MAP_PRIVATE | MAP_FIXED, fd, 0); < WARNING!!!
ASSERT_NE(ptr1, MAP_FAILED);
getchar();
ptr2 = mremap(ptr1, page_size, 2 * page_size,
MREMAP_MAYMOVE | MREMAP_FIXED, ptr1 + 5 * page_size);
ASSERT_NE(ptr2, MAP_FAILED);
ASSERT_NE(mremap(ptr2, page_size, page_size,
MREMAP_MAYMOVE | MREMAP_FIXED, ptr1), MAP_FAILED);
out:
close(fd);
remove(probe_file);
}
The issue originates from the following call chain:
prepare_uprobe
-> copy_insn
-> __copy_insn
-> read_mapping_page() or shmem_read_mapping_page()
Both read_mapping_page() and shmem_read_mapping_page() are invoked
without holding i_rwsem or invalidate_lock. When CONFIG_EXT4_DEBUG
is enabled, this triggers the warning in ext4_check_map_extents_env().
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [BUG] Warning in ext4_check_map_extents_env
2026-03-09 13:41 [BUG] Warning in ext4_check_map_extents_env Pedro Demarchi Gomes
@ 2026-03-10 12:23 ` Zhang Yi
0 siblings, 0 replies; 3+ messages in thread
From: Zhang Yi @ 2026-03-10 12:23 UTC (permalink / raw)
To: Pedro Demarchi Gomes, Theodore Ts'o, Andreas Dilger
Cc: linux-ext4, linux-trace-kernel
Hi,
Thank you for the report!
+ add UPROBES list to CC.
On 3/9/2026 9:41 PM, Pedro Demarchi Gomes wrote:
> Hi,
>
> The warning below occurs when running selftests/mm/merge on a kernel built
> with CONFIG_EXT4_DEBUG=y.
>
> syzkaller login: [ 29.277674] ------------[ cut here ]------------
> [ 29.279636] WARNING: fs/ext4/inode.c:436 at ext4_check_map_extents_env+0x389/0x440, CPU#1: merge/365
> [ 29.282459] Modules linked in:
> [ 29.283497] CPU: 1 UID: 0 PID: 365 Comm: merge Not tainted 7.0.0-rc3 #59 PREEMPT(full)
> [ 29.285999] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
> [ 29.288850] RIP: 0010:ext4_check_map_extents_env+0x389/0x440
> [ 29.290653] Code: ff 48 89 da 48 b8 00 00 00 00 00 fc ff df 48 c1 ea 03 80 3c 02 00 0f 85 87 00 00 00 48 8b 45 18 48 85 c0 0f 85 2e fd ff ff 90 <0f> 0b 90 e9 25 fd ff ff e8 8a 2e ca ff e9 e4 fc ff ff 48 89 df e8
> [ 29.296623] RSP: 0018:ffff88811876f2a8 EFLAGS: 00010246
> [ 29.298288] RAX: 0000000000000000 RBX: ffff888106c66798 RCX: ffffffff81e8ad60
> [ 29.300529] RDX: 1ffff11020d8ccf3 RSI: 0000000000000008 RDI: ffff888106c66798
> [ 29.302748] RBP: ffff888106c66780 R08: 0000000000000000 R09: ffffed1020d8ccf3
> [ 29.305009] R10: ffff888106c6679f R11: 0000000000000001 R12: 00000000001e0c26
> [ 29.307226] R13: ffff88811876f494 R14: ffff88811876f498 R15: ffff88811876f488
> [ 29.309477] FS: 00007fbe82723780(0000) GS:ffff8883c38aa000(0000) knlGS:0000000000000000
> [ 29.312085] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 29.313749] CR2: 00007fbe82790915 CR3: 000000011880e000 CR4: 00000000000006f0
> [ 29.315643] Call Trace:
> [ 29.316170] <TASK>
> [ 29.316539] ext4_map_blocks+0x602/0xb40
> [ 29.317204] ? __pfx_ext4_map_blocks+0x10/0x10
> [ 29.317971] ? __pfx_stack_trace_save+0x10/0x10
> [ 29.318719] ? __filemap_add_folio+0x66f/0xb30
> [ 29.319500] ext4_mpage_readpages.constprop.0.isra.0+0x918/0xe20
> [ 29.320501] ? __pfx_ext4_mpage_readpages.constprop.0.isra.0+0x10/0x10
> [ 29.321602] ? filemap_add_folio+0x11e/0x2c0
> [ 29.322325] ? __pfx_ext4_read_folio+0x10/0x10
> [ 29.323107] ext4_read_folio+0xd9/0x230
> [ 29.323762] ? __pfx_ext4_read_folio+0x10/0x10
> [ 29.324552] ? __pfx_ext4_read_folio+0x10/0x10
> [ 29.325321] filemap_read_folio+0x43/0x150
> [ 29.326041] do_read_cache_folio+0x1b1/0x320
> [ 29.326759] read_cache_page+0x46/0x120
> [ 29.327426] install_breakpoint+0x27d/0x810
It seems this isn't an issue with ext4 itself. Instead, the implementation
in uprobe_mmap() seems not respecting the principle that read_cache_page()
should be called while holding mapping->invalidate_lock.
Although the files tracked through uprobe are generally not modified, so
this should not cause a real stale page cache issue. However, I was
wandering why don't we just follow this rule?
Thanks,
Yi.
> [ 29.328123] uprobe_mmap+0x439/0xeb0
> [ 29.328715] ? __pfx_uprobe_mmap+0x10/0x10
> [ 29.329391] __mmap_region+0x8af/0x2290
> [ 29.330074] ? __pfx___mmap_region+0x10/0x10
> [ 29.330793] ? unwind_get_return_address+0x59/0xa0
> [ 29.331630] ? event_sched_in+0x33f/0xaf0
> [ 29.332308] ? perf_event_groups_next+0x9a/0x200
> [ 29.333667] ? mm_get_unmapped_area_vmflags+0x6c/0xe0
> [ 29.335295] ? mmap_region+0xd3/0x2b0
> [ 29.336171] do_mmap+0x98a/0xec0
> [ 29.336713] ? ipe_mmap_file+0xcf/0xe0
> [ 29.337340] ? __pfx_do_mmap+0x10/0x10
> [ 29.337977] ? __pfx_down_write_killable+0x10/0x10
> [ 29.338757] ? __pfx_mutex_unlock+0x10/0x10
> [ 29.339477] vm_mmap_pgoff+0x1b2/0x310
> [ 29.340114] ? __pfx_vm_mmap_pgoff+0x10/0x10
> [ 29.340819] ? __pfx___do_sys_perf_event_open+0x10/0x10
> [ 29.341688] ? fget+0x189/0x230
> [ 29.342223] ksys_mmap_pgoff+0x2b1/0x5b0
> [ 29.342868] ? __pfx_ksys_mmap_pgoff+0x10/0x10
> [ 29.343618] do_syscall_64+0xe2/0x560
> [ 29.344246] entry_SYSCALL_64_after_hwframe+0x77/0x7f
> [ 29.345105] RIP: 0033:0x7fbe82832de2
> [ 29.345707] Code: 00 00 00 0f 1f 44 00 00 41 f7 c1 ff 0f 00 00 75 27 55 89 cd 53 48 89 fb 48 85 ff 74 3b 41 89 ea 48 89 df b8 09 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 76 5b 5d c3 0f 1f 00 48 8b 05 f9 8f 0d 00 64
> [ 29.348851] RSP: 002b:00007ffd535f3b38 EFLAGS: 00000206 ORIG_RAX: 0000000000000009
> [ 29.350309] RAX: ffffffffffffffda RBX: 00007fbe826f9000 RCX: 00007fbe82832de2
> [ 29.351526] RDX: 0000000000000004 RSI: 000000000000a000 RDI: 00007fbe826f9000
> [ 29.352865] RBP: 0000000000000012 R08: 000000000000000b R09: 0000000000000000
> [ 29.354136] R10: 0000000000000012 R11: 0000000000000206 R12: 00007ffd535f4120
> [ 29.355357] R13: 0000000000000000 R14: 000000000000000b R15: 0000000000001000
> [ 29.356684] </TASK>
> [ 29.357119] ---[ end trace 0000000000000000 ]---
>
> The warning is triggered by TEST_F(merge, handle_uprobe_upon_merged_vma) in
> selftests/mm/merge.c. It occurs during the first mmap() following the
> __NR_perf_event_open syscall, as shown in the excerpt below:
>
> TEST_F(merge, handle_uprobe_upon_merged_vma)
> {
> const size_t attr_sz = sizeof(struct perf_event_attr);
> unsigned int page_size = self->page_size;
> const char *probe_file = "./foo";
> char *carveout = self->carveout;
> struct perf_event_attr attr;
> unsigned long type;
> void *ptr1, *ptr2;
> int fd;
>
> fd = open(probe_file, O_RDWR|O_CREAT, 0600);
> ASSERT_GE(fd, 0);
>
> ASSERT_EQ(ftruncate(fd, page_size), 0);
> if (read_sysfs("/sys/bus/event_source/devices/uprobe/type", &type) != 0) {
> SKIP(goto out, "Failed to read uprobe sysfs file, skipping");
> }
>
> memset(&attr, 0, attr_sz);
> attr.size = attr_sz;
> attr.type = type;
> attr.config1 = (__u64)(long)probe_file;
> attr.config2 = 0x0;
>
> ASSERT_GE(syscall(__NR_perf_event_open, &attr, 0, -1, -1, 0), 0);
>
> ptr1 = mmap(&carveout[page_size], 10 * page_size, PROT_EXEC,
> MAP_PRIVATE | MAP_FIXED, fd, 0); < WARNING!!!
> ASSERT_NE(ptr1, MAP_FAILED);
>
> getchar();
>
> ptr2 = mremap(ptr1, page_size, 2 * page_size,
> MREMAP_MAYMOVE | MREMAP_FIXED, ptr1 + 5 * page_size);
> ASSERT_NE(ptr2, MAP_FAILED);
>
> ASSERT_NE(mremap(ptr2, page_size, page_size,
> MREMAP_MAYMOVE | MREMAP_FIXED, ptr1), MAP_FAILED);
>
> out:
> close(fd);
> remove(probe_file);
> }
>
> The issue originates from the following call chain:
> prepare_uprobe
> -> copy_insn
> -> __copy_insn
> -> read_mapping_page() or shmem_read_mapping_page()
>
> Both read_mapping_page() and shmem_read_mapping_page() are invoked
> without holding i_rwsem or invalidate_lock. When CONFIG_EXT4_DEBUG
> is enabled, this triggers the warning in ext4_check_map_extents_env().
>
>
>
^ permalink raw reply [flat|nested] 3+ messages in thread
* [BUG] Warning in ext4_check_map_extents_env
@ 2026-03-09 4:26 Pedro Demarchi Gomes
0 siblings, 0 replies; 3+ messages in thread
From: Pedro Demarchi Gomes @ 2026-03-09 4:26 UTC (permalink / raw)
To: linux-ext4
Hi,
The warning below occurs when running selftests/mm/merge on a kernel built
with CONFIG_EXT4_DEBUG=y.
syzkaller login: [ 29.277674] ------------[ cut here ]------------
[ 29.279636] WARNING: fs/ext4/inode.c:436 at ext4_check_map_extents_env+0x389/0x440, CPU#1: merge/365
[ 29.282459] Modules linked in:
[ 29.283497] CPU: 1 UID: 0 PID: 365 Comm: merge Not tainted 7.0.0-rc3 #59 PREEMPT(full)
[ 29.285999] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[ 29.288850] RIP: 0010:ext4_check_map_extents_env+0x389/0x440
[ 29.290653] Code: ff 48 89 da 48 b8 00 00 00 00 00 fc ff df 48 c1 ea 03 80 3c 02 00 0f 85 87 00 00 00 48 8b 45 18 48 85 c0 0f 85 2e fd ff ff 90 <0f> 0b 90 e9 25 fd ff ff e8 8a 2e ca ff e9 e4 fc ff ff 48 89 df e8
[ 29.296623] RSP: 0018:ffff88811876f2a8 EFLAGS: 00010246
[ 29.298288] RAX: 0000000000000000 RBX: ffff888106c66798 RCX: ffffffff81e8ad60
[ 29.300529] RDX: 1ffff11020d8ccf3 RSI: 0000000000000008 RDI: ffff888106c66798
[ 29.302748] RBP: ffff888106c66780 R08: 0000000000000000 R09: ffffed1020d8ccf3
[ 29.305009] R10: ffff888106c6679f R11: 0000000000000001 R12: 00000000001e0c26
[ 29.307226] R13: ffff88811876f494 R14: ffff88811876f498 R15: ffff88811876f488
[ 29.309477] FS: 00007fbe82723780(0000) GS:ffff8883c38aa000(0000) knlGS:0000000000000000
[ 29.312085] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 29.313749] CR2: 00007fbe82790915 CR3: 000000011880e000 CR4: 00000000000006f0
[ 29.315643] Call Trace:
[ 29.316170] <TASK>
[ 29.316539] ext4_map_blocks+0x602/0xb40
[ 29.317204] ? __pfx_ext4_map_blocks+0x10/0x10
[ 29.317971] ? __pfx_stack_trace_save+0x10/0x10
[ 29.318719] ? __filemap_add_folio+0x66f/0xb30
[ 29.319500] ext4_mpage_readpages.constprop.0.isra.0+0x918/0xe20
[ 29.320501] ? __pfx_ext4_mpage_readpages.constprop.0.isra.0+0x10/0x10
[ 29.321602] ? filemap_add_folio+0x11e/0x2c0
[ 29.322325] ? __pfx_ext4_read_folio+0x10/0x10
[ 29.323107] ext4_read_folio+0xd9/0x230
[ 29.323762] ? __pfx_ext4_read_folio+0x10/0x10
[ 29.324552] ? __pfx_ext4_read_folio+0x10/0x10
[ 29.325321] filemap_read_folio+0x43/0x150
[ 29.326041] do_read_cache_folio+0x1b1/0x320
[ 29.326759] read_cache_page+0x46/0x120
[ 29.327426] install_breakpoint+0x27d/0x810
[ 29.328123] uprobe_mmap+0x439/0xeb0
[ 29.328715] ? __pfx_uprobe_mmap+0x10/0x10
[ 29.329391] __mmap_region+0x8af/0x2290
[ 29.330074] ? __pfx___mmap_region+0x10/0x10
[ 29.330793] ? unwind_get_return_address+0x59/0xa0
[ 29.331630] ? event_sched_in+0x33f/0xaf0
[ 29.332308] ? perf_event_groups_next+0x9a/0x200
[ 29.333667] ? mm_get_unmapped_area_vmflags+0x6c/0xe0
[ 29.335295] ? mmap_region+0xd3/0x2b0
[ 29.336171] do_mmap+0x98a/0xec0
[ 29.336713] ? ipe_mmap_file+0xcf/0xe0
[ 29.337340] ? __pfx_do_mmap+0x10/0x10
[ 29.337977] ? __pfx_down_write_killable+0x10/0x10
[ 29.338757] ? __pfx_mutex_unlock+0x10/0x10
[ 29.339477] vm_mmap_pgoff+0x1b2/0x310
[ 29.340114] ? __pfx_vm_mmap_pgoff+0x10/0x10
[ 29.340819] ? __pfx___do_sys_perf_event_open+0x10/0x10
[ 29.341688] ? fget+0x189/0x230
[ 29.342223] ksys_mmap_pgoff+0x2b1/0x5b0
[ 29.342868] ? __pfx_ksys_mmap_pgoff+0x10/0x10
[ 29.343618] do_syscall_64+0xe2/0x560
[ 29.344246] entry_SYSCALL_64_after_hwframe+0x77/0x7f
[ 29.345105] RIP: 0033:0x7fbe82832de2
[ 29.345707] Code: 00 00 00 0f 1f 44 00 00 41 f7 c1 ff 0f 00 00 75 27 55 89 cd 53 48 89 fb 48 85 ff 74 3b 41 89 ea 48 89 df b8 09 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 76 5b 5d c3 0f 1f 00 48 8b 05 f9 8f 0d 00 64
[ 29.348851] RSP: 002b:00007ffd535f3b38 EFLAGS: 00000206 ORIG_RAX: 0000000000000009
[ 29.350309] RAX: ffffffffffffffda RBX: 00007fbe826f9000 RCX: 00007fbe82832de2
[ 29.351526] RDX: 0000000000000004 RSI: 000000000000a000 RDI: 00007fbe826f9000
[ 29.352865] RBP: 0000000000000012 R08: 000000000000000b R09: 0000000000000000
[ 29.354136] R10: 0000000000000012 R11: 0000000000000206 R12: 00007ffd535f4120
[ 29.355357] R13: 0000000000000000 R14: 000000000000000b R15: 0000000000001000
[ 29.356684] </TASK>
[ 29.357119] ---[ end trace 0000000000000000 ]---
The warning is triggered by TEST_F(merge, handle_uprobe_upon_merged_vma) in
selftests/mm/merge.c. It occurs during the first mmap() following the
__NR_perf_event_open syscall, as shown in the excerpt below:
TEST_F(merge, handle_uprobe_upon_merged_vma)
{
const size_t attr_sz = sizeof(struct perf_event_attr);
unsigned int page_size = self->page_size;
const char *probe_file = "./foo";
char *carveout = self->carveout;
struct perf_event_attr attr;
unsigned long type;
void *ptr1, *ptr2;
int fd;
fd = open(probe_file, O_RDWR|O_CREAT, 0600);
ASSERT_GE(fd, 0);
ASSERT_EQ(ftruncate(fd, page_size), 0);
if (read_sysfs("/sys/bus/event_source/devices/uprobe/type", &type) != 0) {
SKIP(goto out, "Failed to read uprobe sysfs file, skipping");
}
memset(&attr, 0, attr_sz);
attr.size = attr_sz;
attr.type = type;
attr.config1 = (__u64)(long)probe_file;
attr.config2 = 0x0;
ASSERT_GE(syscall(__NR_perf_event_open, &attr, 0, -1, -1, 0), 0);
ptr1 = mmap(&carveout[page_size], 10 * page_size, PROT_EXEC,
MAP_PRIVATE | MAP_FIXED, fd, 0); < WARNING!!!
ASSERT_NE(ptr1, MAP_FAILED);
getchar();
ptr2 = mremap(ptr1, page_size, 2 * page_size,
MREMAP_MAYMOVE | MREMAP_FIXED, ptr1 + 5 * page_size);
ASSERT_NE(ptr2, MAP_FAILED);
ASSERT_NE(mremap(ptr2, page_size, page_size,
MREMAP_MAYMOVE | MREMAP_FIXED, ptr1), MAP_FAILED);
out:
close(fd);
remove(probe_file);
}
The issue originates from the following call chain:
prepare_uprobe
-> copy_insn
-> __copy_insn
-> read_mapping_page() or shmem_read_mapping_page()
Both read_mapping_page() and shmem_read_mapping_page() are invoked
without holding i_rwsem or invalidate_lock. When CONFIG_EXT4_DEBUG
is enabled, this triggers the warning in ext4_check_map_extents_env().
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-03-10 12:23 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-03-09 13:41 [BUG] Warning in ext4_check_map_extents_env Pedro Demarchi Gomes
2026-03-10 12:23 ` Zhang Yi
-- strict thread matches above, loose matches on Subject: below --
2026-03-09 4:26 Pedro Demarchi Gomes
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox