gcov: Protect from uninitialized number of functions provided by GCC

RFC PATCH In review Leon Romanovsky 4 participants 5 messages Aug 27, 2020 – Aug 30, 2020

Subsystems (2)

Memory management Gcov based kernel profiling

lore

Share & tools

lore

Leon Romanovsky

author

2020-08-27 13:39 UTC #0 lore raw

RFC PATCH 1/1 gcov: Protect from uninitialized number of functions provided by GCC

From: Leon Romanovsky <[email protected]>

The kernel compiled with GCC 10.2.1 and KASAN together with GCOV enabled produces the following splats while reloading modules.

First splat [1] is generated due to the situation that gcov_info can be both user and kernel pointer, the memcpy() during kmemdup() causes to this. As a possible solution copy fields manually.

Second splat [2] is seen because n_function provided by GCC through __gcov_init() is ridiculously high, in my case it was 2698213824. IMHO it means that this field is not initialized, but I'm not sure.

[1]
==================================================================

 BUG: KASAN: global-out-of-bounds in kmemdup+0x43/0x70
 Read of size 120 at addr ffffffffa0d2c780 by task modprobe/296

 CPU: 0 PID: 296 Comm: modprobe Not tainted 5.9.0-rc1+ #1860
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04 /01/2014
 Call Trace:
  ? dump_stack+0x128/0x1af
  ? print_address_description.constprop.0+0x2c/0x3f0
  ? _raw_spin_lock_irqsave+0x34/0xa0
  ? __kasan_check_read+0x1d/0x30
  ? kmemdup+0x43/0x70
  ? kmemdup+0x43/0x70
  ? gcov_info_dup+0x2d/0x730
  ? __kasan_check_write+0x20/0x30
  ? __mutex_unlock_slowpath+0x10d/0x740
  ? gcov_event+0x88d/0xd30
  ? gcov_module_notifier+0xe9/0x100
  ? notifier_call_chain+0xeb/0x170
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? do_init_module+0x810/0x810
  ? syscall_enter_from_user_mode+0x40/0x420
  ? trace_hardirqs_on+0x45/0xb0
  ? syscall_enter_from_user_mode+0x40/0x420
  ? do_syscall_64+0x45/0x70
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9

 The buggy address belongs to the variable:
  __gcov_.uverbs_attr_get_obj+0x60/0xfffffffffff778e0 [mlx5_ib]

 Memory state around the buggy address:
  ffffffffa0d2c680: 00 f9 f9 f9 f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9
  ffffffffa0d2c700: f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9 f9 f9 f9 f9
 >ffffffffa0d2c780: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 f9 f9
                                                              ^
  ffffffffa0d2c800: f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9 f9 f9 f9 f9
  ffffffffa0d2c880: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 ==================================================================
 Disabling lock debugging due to kernel taint
 gcov: could not save data for '/home/leonro/src/kernel/drivers/infiniband/hw/mlx5/std_types.gcda' (out o

f memory)

[2]
Colin has similar error [3].

 ------------[ cut here ]------------
 WARNING: CPU: 0 PID: 296 at mm/page_alloc.c:4859 __alloc_pages_nodemask+0x670/0x3190
 Modules linked in: mlx5_ib(-) mlx5_core mlxfw ptp ib_ipoib pps_core rdma_ucm rdma_cm iw_cm ib_cm ib_umad  ib_uverbs ib_core
 CPU: 0 PID: 296 Comm: modprobe Tainted: G    B             5.9.0-rc1+ #1860
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04 /01/2014
 RIP: 0010:__alloc_pages_nodemask+0x670/0x3190
 Code: e9 af fc ff ff 48 83 05 fd 28 90 05 01 81 e7 00 20 00 00 48 c7 44 24 28 00 00 00 00 0f 85 fb fd ff  ff 48 83 05 f0 28 90 05 01 <0f> 0b 48 83 05 ee 28 90 05 01 48 83 05 ee 28 90 05 01 e9 dc fd ff
 RSP: 0018:ffff88805f7ffa28 EFLAGS: 00010202
 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 1ffff1100befff5e
 RDX: 0000000000000000 RSI: 0000000000000017 RDI: 0000000000000000
 RBP: 000000050695a900 R08: ffff888060fc7900 R09: ffff888060fc793b
 R10: ffffed100c1f8f27 R11: ffffed100c1f8f28 R12: 0000000000040dc0
 R13: 000000050695a900 R14: 0000000000000017 R15: 0000000000000001
 FS:  00007f521f695740(0000) GS:ffff88806ce00000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00007f31b013f000 CR3: 000000006637e001 CR4: 0000000000370eb0
 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
 Call Trace:
  ? __kmalloc_track_caller+0x17a/0x570
  ? gcov_info_dup+0xfe/0x730
  ? gcov_event+0x88d/0xd30
  ? gcov_module_notifier+0xe9/0x100
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9
  ? mark_lock+0xba0/0xba0
  ? mark_lock+0xba0/0xba0
  ? notifier_call_chain+0xeb/0x170
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? do_syscall_64+0x45/0x70
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9
  ? warn_alloc+0x130/0x130
  ? lock_acquire+0x1f2/0xa30
  ? fs_reclaim_acquire+0x1f/0x70
  ? fs_reclaim_release+0x1f/0x50
  ? __kasan_check_read+0x1d/0x30
  ? reacquire_held_locks+0x420/0x420
  ? reacquire_held_locks+0x420/0x420
  kmalloc_order+0x3f/0xc0
  kmalloc_order_trace+0x24/0x220
  __kmalloc+0x41b/0x5a0
  ? gcov_info_dup+0xfe/0x730
  ? memcpy+0x73/0xa0
  gcov_info_dup+0x176/0x730
  gcov_event+0x88d/0xd30
  gcov_module_notifier+0xe9/0x100
  notifier_call_chain+0xeb/0x170
  blocking_notifier_call_chain+0x75/0xc0
  __x64_sys_delete_module+0x326/0x5a0
  ? do_init_module+0x810/0x810
  ? syscall_enter_from_user_mode+0x40/0x420
  ? trace_hardirqs_on+0x45/0xb0
  ? syscall_enter_from_user_mode+0x40/0x420
  do_syscall_64+0x45/0x70
  entry_SYSCALL_64_after_hwframe+0x44/0xa9
 RIP: 0033:0x7f521f7c531b
 Code: 73 01 c3 48 8b 0d 7d 0b 0c 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa b8 b0 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 4d 0b 0c 00 f7 d8 64 89 01 48
 RSP: 002b:00007ffe1bd4af48 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0
 RAX: ffffffffffffffda RBX: 0000561a3eae0910 RCX: 00007f521f7c531b
 RDX: 0000000000000000 RSI: 0000000000000800 RDI: 0000561a3eae0978
 RBP: 0000561a3eae0910 R08: 1999999999999999 R09: 0000000000000000
 R10: 00007f521f839ac0 R11: 0000000000000206 R12: 0000000000000000
 R13: 0000561a3eae0978 R14: 0000000000000000 R15: 0000561a3eae84d0
 irq event stamp: 326464
 hardirqs last  enabled at (326463): [<ffffffff832ecdde>] _raw_spin_unlock_irqrestore+0x8e/0xb0
 hardirqs last disabled at (326464): [<ffffffff832ec994>] _raw_spin_lock_irqsave+0x34/0xa0
 hardirqs last disabled at (326464): [<ffffffff832ec994>] _raw_spin_lock_irqsave+0x34/0xa0
 softirqs last  enabled at (320794): [<ffffffff83600931>] __do_softirq+0x931/0xbc4
 softirqs last disabled at (320789): [<ffffffff83400f2f>] asm_call_on_stack+0xf/0x20
 ---[ end trace 065ea9cc2ba144a6 ]---

[3] https://bugzilla.kernel.org/show_bug.cgi?id=208885#c1

Cc Colin Ian King <[email protected]>

Signed-off-by Leon Romanovsky <[email protected]>

---
I have a strong feeling that this solution is not correct, but don't know how to do it right. The problem exists and reproducable in seconds.

---
 kernel/gcov/gcc_4_7.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)

kernel/gcov/gcc_4_7.c +11 -8

index 908fdf5098c3..357ef839cdd3 100644
@@ -275,20 +275,23 @@ struct gcov_info *gcov_info_dup(struct gcov_info *info)
 	size_t fi_size; /* function info size */
 	size_t cv_size; /* counter values size */
 
-	dup = kmemdup(info, sizeof(*dup), GFP_KERNEL);
+	dup = kzalloc(sizeof(*dup), GFP_KERNEL);
 	if (!dup)
 		return NULL;
 
-	dup->next = NULL;
-	dup->filename = NULL;
-	dup->functions = NULL;
+	dup->version = info->version;
+	dup->stamp = info->stamp;
+	for (fi_idx = 0; i < GCOV_COUNTERS; i++)
+		dup->merge[i] = info->merge[i];
+	dup->n_functions = info->n_functions;
 
-	dup->filename = kstrdup(info->filename, GFP_KERNEL);
+	dup->filename = kstrdup_const(info->filename, GFP_KERNEL);
 	if (!dup->filename)
 		goto err_free;
 
-	dup->functions = kcalloc(info->n_functions,
-				 sizeof(struct gcov_fn_info *), GFP_KERNEL);
+	dup->functions =
+		kcalloc(info->n_functions, sizeof(struct gcov_fn_info *),
+			GFP_KERNEL | __GFP_NOWARN);
 	if (!dup->functions)
 		goto err_free;

@@ -359,7 +362,7 @@ void gcov_info_free(struct gcov_info *info)
 
 free_info:
 	kfree(info->functions);
-	kfree(info->filename);
+	kfree_const(info->filename);
 	kfree(info);
 }

--
2.26.2

Leon Romanovsky

2020-08-27 15:13 UTC #1 lore raw

Quoted 153 lines · Leon Romanovsky

From: Leon Romanovsky <[email protected]>

The kernel compiled with GCC 10.2.1 and KASAN together with GCOV enabled produces the following splats while reloading modules.

First splat [1] is generated due to the situation that gcov_info can be both user and kernel pointer, the memcpy() during kmemdup() causes to this. As a possible solution copy fields manually.

[1]
==================================================================

 BUG: KASAN: global-out-of-bounds in kmemdup+0x43/0x70
 Read of size 120 at addr ffffffffa0d2c780 by task modprobe/296

 CPU: 0 PID: 296 Comm: modprobe Not tainted 5.9.0-rc1+ #1860
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04 /01/2014
 Call Trace:
  ? dump_stack+0x128/0x1af
  ? print_address_description.constprop.0+0x2c/0x3f0
  ? _raw_spin_lock_irqsave+0x34/0xa0
  ? __kasan_check_read+0x1d/0x30
  ? kmemdup+0x43/0x70
  ? kmemdup+0x43/0x70
  ? gcov_info_dup+0x2d/0x730
  ? __kasan_check_write+0x20/0x30
  ? __mutex_unlock_slowpath+0x10d/0x740
  ? gcov_event+0x88d/0xd30
  ? gcov_module_notifier+0xe9/0x100
  ? notifier_call_chain+0xeb/0x170
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? do_init_module+0x810/0x810
  ? syscall_enter_from_user_mode+0x40/0x420
  ? trace_hardirqs_on+0x45/0xb0
  ? syscall_enter_from_user_mode+0x40/0x420
  ? do_syscall_64+0x45/0x70
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9

 The buggy address belongs to the variable:
  __gcov_.uverbs_attr_get_obj+0x60/0xfffffffffff778e0 [mlx5_ib]

 Memory state around the buggy address:
  ffffffffa0d2c680: 00 f9 f9 f9 f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9
  ffffffffa0d2c700: f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9 f9 f9 f9 f9
 >ffffffffa0d2c780: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 f9 f9
                                                              ^
  ffffffffa0d2c800: f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9 f9 f9 f9 f9
  ffffffffa0d2c880: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 ==================================================================
 Disabling lock debugging due to kernel taint
 gcov: could not save data for '/home/leonro/src/kernel/drivers/infiniband/hw/mlx5/std_types.gcda' (out o

f memory)

[2]
Colin has similar error [3].

 ------------[ cut here ]------------
 WARNING: CPU: 0 PID: 296 at mm/page_alloc.c:4859 __alloc_pages_nodemask+0x670/0x3190
 Modules linked in: mlx5_ib(-) mlx5_core mlxfw ptp ib_ipoib pps_core rdma_ucm rdma_cm iw_cm ib_cm ib_umad  ib_uverbs ib_core
 CPU: 0 PID: 296 Comm: modprobe Tainted: G    B             5.9.0-rc1+ #1860
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04 /01/2014
 RIP: 0010:__alloc_pages_nodemask+0x670/0x3190
 Code: e9 af fc ff ff 48 83 05 fd 28 90 05 01 81 e7 00 20 00 00 48 c7 44 24 28 00 00 00 00 0f 85 fb fd ff  ff 48 83 05 f0 28 90 05 01 <0f> 0b 48 83 05 ee 28 90 05 01 48 83 05 ee 28 90 05 01 e9 dc fd ff
 RSP: 0018:ffff88805f7ffa28 EFLAGS: 00010202
 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 1ffff1100befff5e
 RDX: 0000000000000000 RSI: 0000000000000017 RDI: 0000000000000000
 RBP: 000000050695a900 R08: ffff888060fc7900 R09: ffff888060fc793b
 R10: ffffed100c1f8f27 R11: ffffed100c1f8f28 R12: 0000000000040dc0
 R13: 000000050695a900 R14: 0000000000000017 R15: 0000000000000001
 FS:  00007f521f695740(0000) GS:ffff88806ce00000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00007f31b013f000 CR3: 000000006637e001 CR4: 0000000000370eb0
 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
 Call Trace:
  ? __kmalloc_track_caller+0x17a/0x570
  ? gcov_info_dup+0xfe/0x730
  ? gcov_event+0x88d/0xd30
  ? gcov_module_notifier+0xe9/0x100
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9
  ? mark_lock+0xba0/0xba0
  ? mark_lock+0xba0/0xba0
  ? notifier_call_chain+0xeb/0x170
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? do_syscall_64+0x45/0x70
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9
  ? warn_alloc+0x130/0x130
  ? lock_acquire+0x1f2/0xa30
  ? fs_reclaim_acquire+0x1f/0x70
  ? fs_reclaim_release+0x1f/0x50
  ? __kasan_check_read+0x1d/0x30
  ? reacquire_held_locks+0x420/0x420
  ? reacquire_held_locks+0x420/0x420
  kmalloc_order+0x3f/0xc0
  kmalloc_order_trace+0x24/0x220
  __kmalloc+0x41b/0x5a0
  ? gcov_info_dup+0xfe/0x730
  ? memcpy+0x73/0xa0
  gcov_info_dup+0x176/0x730
  gcov_event+0x88d/0xd30
  gcov_module_notifier+0xe9/0x100
  notifier_call_chain+0xeb/0x170
  blocking_notifier_call_chain+0x75/0xc0
  __x64_sys_delete_module+0x326/0x5a0
  ? do_init_module+0x810/0x810
  ? syscall_enter_from_user_mode+0x40/0x420
  ? trace_hardirqs_on+0x45/0xb0
  ? syscall_enter_from_user_mode+0x40/0x420
  do_syscall_64+0x45/0x70
  entry_SYSCALL_64_after_hwframe+0x44/0xa9
 RIP: 0033:0x7f521f7c531b
 Code: 73 01 c3 48 8b 0d 7d 0b 0c 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa b8 b0 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 4d 0b 0c 00 f7 d8 64 89 01 48
 RSP: 002b:00007ffe1bd4af48 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0
 RAX: ffffffffffffffda RBX: 0000561a3eae0910 RCX: 00007f521f7c531b
 RDX: 0000000000000000 RSI: 0000000000000800 RDI: 0000561a3eae0978
 RBP: 0000561a3eae0910 R08: 1999999999999999 R09: 0000000000000000
 R10: 00007f521f839ac0 R11: 0000000000000206 R12: 0000000000000000
 R13: 0000561a3eae0978 R14: 0000000000000000 R15: 0000561a3eae84d0
 irq event stamp: 326464
 hardirqs last  enabled at (326463): [<ffffffff832ecdde>] _raw_spin_unlock_irqrestore+0x8e/0xb0
 hardirqs last disabled at (326464): [<ffffffff832ec994>] _raw_spin_lock_irqsave+0x34/0xa0
 hardirqs last disabled at (326464): [<ffffffff832ec994>] _raw_spin_lock_irqsave+0x34/0xa0
 softirqs last  enabled at (320794): [<ffffffff83600931>] __do_softirq+0x931/0xbc4
 softirqs last disabled at (320789): [<ffffffff83400f2f>] asm_call_on_stack+0xf/0x20
 ---[ end trace 065ea9cc2ba144a6 ]---

[3] https://bugzilla.kernel.org/show_bug.cgi?id=208885#c1

Cc Colin Ian King <[email protected]>

Signed-off-by Leon Romanovsky <[email protected]>

---
I have a strong feeling that this solution is not correct, but don't know how to do it right. The problem exists and reproducable in seconds.

---
 kernel/gcov/gcc_4_7.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)

kernel/gcov/gcc_4_7.c +1 -1

index 908fdf5098c3..357ef839cdd3 100644
@@ -275,20 +275,23 @@ struct gcov_info *gcov_info_dup(struct gcov_info *info)
 	size_t fi_size; /* function info size */
 	size_t cv_size; /* counter values size */
 
-	dup = kmemdup(info, sizeof(*dup), GFP_KERNEL);
+	dup = kzalloc(sizeof(*dup), GFP_KERNEL);

kernel/gcov/gcc_4_7.c struct gcov_info *gcov_info_dup(struct gcov_info *info) · line 285

Quoted 10 lines

kernel/gcov/gcc_4_7.c +4 -3

 	if (!dup)
 		return NULL;
 
-	dup->next = NULL;
-	dup->filename = NULL;
-	dup->functions = NULL;
+	dup->version = info->version;
+	dup->stamp = info->stamp;
+	for (fi_idx = 0; i < GCOV_COUNTERS; i++)
+		dup->merge[i] = info->merge[i];

And of course "i" should be replaced to be "fi_idx". But I'm confident that the solution is not right anyway.

Thanks

kernel test robot

build bot

2020-08-27 21:08 UTC #2 lore raw

In reply to Leon Romanovsky

[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in https://git-scm.com/docs/git-format-patch]

Hi Leon,

[FYI, it's a private test report for your RFC patch.]
[auto build test ERROR on hnaz-linux-mm/master]
[also build test ERROR on linux/master linus/master v5.9-rc2 next-20200827] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Leon-Romanovsky/gcov-Protect-from-uninitialized-number-of-functions-provided-by-GCC/20200827-214008
base:   https://github.com/hnaz/linux-mm master
config: m68k-randconfig-s032-20200827 (attached as .config)
compiler: m68k-linux-gcc (GCC) 9.3.0

reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# apt-get install sparse
# sparse version: v0.6.2-191-g10164920-dirty
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' ARCH=m68k

If you fix the issue, kindly add following tag as appropriate

Reported-by kernel test robot <[email protected]>

All errors (new ones prefixed by >>):

kernel/gcov/gcc_4_7.c: In function 'gcov_info_dup':

>> kernel/gcov/gcc_4_7.c:284:19: error: 'i' undeclared (first use in this function)
     284 |  for (fi_idx = 0; i < GCOV_COUNTERS; i++)
         |                   ^

kernel/gcov/gcc_4_7.c:284:19: note: each undeclared identifier is reported only once for each function it appears in

# https://github.com/0day-ci/linux/commit/9087d4a9ebf3be8ae595ecef237b57270d78b4cb
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Leon-Romanovsky/gcov-Protect-from-uninitialized-number-of-functions-provided-by-GCC/20200827-214008 git checkout 9087d4a9ebf3be8ae595ecef237b57270d78b4cb
vim +/i +284 kernel/gcov/gcc_4_7.c

260
261 /**
262 * gcov_info_dup - duplicate profiling data set
263 * @info: profiling data set to duplicate
264 *
265 * Return newly allocated duplicate on success, %NULL on error.
266 */
267 struct gcov_info *gcov_info_dup(struct gcov_info *info)
268 {
269 struct gcov_info *dup;
270 struct gcov_ctr_info *dci_ptr; /* dst counter info */
271 struct gcov_ctr_info *sci_ptr; /* src counter info */
272 unsigned int active;
273 unsigned int fi_idx; /* function info idx */
274 unsigned int ct_idx; /* counter type idx */
275 size_t fi_size; /* function info size */
276 size_t cv_size; /* counter values size */
277
278 dup = kzalloc(sizeof(*dup), GFP_KERNEL);
279 if (!dup)
280 return NULL;
281
282 dup->version = info->version;
283 dup->stamp = info->stamp;

 > 284		for (fi_idx = 0; i < GCOV_COUNTERS; i++)

285 dup->merge[i] = info->merge[i];
286 dup->n_functions = info->n_functions;
287
288 dup->filename = kstrdup_const(info->filename, GFP_KERNEL);
289 if (!dup->filename)
290 goto err_free;
291
292 dup->functions =
293 kcalloc(info->n_functions, sizeof(struct gcov_fn_info *),
294 GFP_KERNEL | __GFP_NOWARN);
295 if (!dup->functions)
296 goto err_free;
297
298 active = num_counter_active(info);
299 fi_size = sizeof(struct gcov_fn_info);
300 fi_size += sizeof(struct gcov_ctr_info) * active;
301
302 for (fi_idx = 0; fi_idx < info->n_functions; fi_idx++) {
303 dup->functions[fi_idx] = kzalloc(fi_size, GFP_KERNEL);
304 if (!dup->functions[fi_idx])
305 goto err_free;
306
307 *(dup->functions[fi_idx]) = *(info->functions[fi_idx]);
308
309 sci_ptr = info->functions[fi_idx]->ctrs;
310 dci_ptr = dup->functions[fi_idx]->ctrs;
311
312 for (ct_idx = 0; ct_idx < active; ct_idx++) {
313
314 cv_size = sizeof(gcov_type) * sci_ptr->num;
315
316 dci_ptr->values = vmalloc(cv_size);
317
318 if (!dci_ptr->values)
319 goto err_free;
320
321 dci_ptr->num = sci_ptr->num;
322 memcpy(dci_ptr->values, sci_ptr->values, cv_size);
323
324 sci_ptr++;
325 dci_ptr++;
326 }
327 }
328
329 return dup;
330 err_free:
331 gcov_info_free(dup);
332 return NULL;
333 }
334

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org

Colin Ian King

2020-08-29 14:12 UTC #3 lore raw

In reply to Leon Romanovsky

Quoted 152 lines · Leon Romanovsky

From: Leon Romanovsky <[email protected]>

The kernel compiled with GCC 10.2.1 and KASAN together with GCOV enabled produces the following splats while reloading modules.

First splat [1] is generated due to the situation that gcov_info can be both user and kernel pointer, the memcpy() during kmemdup() causes to this. As a possible solution copy fields manually.

[1]
==================================================================

 BUG: KASAN: global-out-of-bounds in kmemdup+0x43/0x70
 Read of size 120 at addr ffffffffa0d2c780 by task modprobe/296

 CPU: 0 PID: 296 Comm: modprobe Not tainted 5.9.0-rc1+ #1860
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04 /01/2014
 Call Trace:
  ? dump_stack+0x128/0x1af
  ? print_address_description.constprop.0+0x2c/0x3f0
  ? _raw_spin_lock_irqsave+0x34/0xa0
  ? __kasan_check_read+0x1d/0x30
  ? kmemdup+0x43/0x70
  ? kmemdup+0x43/0x70
  ? gcov_info_dup+0x2d/0x730
  ? __kasan_check_write+0x20/0x30
  ? __mutex_unlock_slowpath+0x10d/0x740
  ? gcov_event+0x88d/0xd30
  ? gcov_module_notifier+0xe9/0x100
  ? notifier_call_chain+0xeb/0x170
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? do_init_module+0x810/0x810
  ? syscall_enter_from_user_mode+0x40/0x420
  ? trace_hardirqs_on+0x45/0xb0
  ? syscall_enter_from_user_mode+0x40/0x420
  ? do_syscall_64+0x45/0x70
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9

 The buggy address belongs to the variable:
  __gcov_.uverbs_attr_get_obj+0x60/0xfffffffffff778e0 [mlx5_ib]

 Memory state around the buggy address:
  ffffffffa0d2c680: 00 f9 f9 f9 f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9
  ffffffffa0d2c700: f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9 f9 f9 f9 f9
 >ffffffffa0d2c780: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 f9 f9
                                                              ^
  ffffffffa0d2c800: f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9 f9 f9 f9 f9
  ffffffffa0d2c880: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 ==================================================================
 Disabling lock debugging due to kernel taint
 gcov: could not save data for '/home/leonro/src/kernel/drivers/infiniband/hw/mlx5/std_types.gcda' (out o

f memory)

[2]
Colin has similar error [3].

 ------------[ cut here ]------------
 WARNING: CPU: 0 PID: 296 at mm/page_alloc.c:4859 __alloc_pages_nodemask+0x670/0x3190
 Modules linked in: mlx5_ib(-) mlx5_core mlxfw ptp ib_ipoib pps_core rdma_ucm rdma_cm iw_cm ib_cm ib_umad  ib_uverbs ib_core
 CPU: 0 PID: 296 Comm: modprobe Tainted: G    B             5.9.0-rc1+ #1860
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04 /01/2014
 RIP: 0010:__alloc_pages_nodemask+0x670/0x3190
 Code: e9 af fc ff ff 48 83 05 fd 28 90 05 01 81 e7 00 20 00 00 48 c7 44 24 28 00 00 00 00 0f 85 fb fd ff  ff 48 83 05 f0 28 90 05 01 <0f> 0b 48 83 05 ee 28 90 05 01 48 83 05 ee 28 90 05 01 e9 dc fd ff
 RSP: 0018:ffff88805f7ffa28 EFLAGS: 00010202
 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 1ffff1100befff5e
 RDX: 0000000000000000 RSI: 0000000000000017 RDI: 0000000000000000
 RBP: 000000050695a900 R08: ffff888060fc7900 R09: ffff888060fc793b
 R10: ffffed100c1f8f27 R11: ffffed100c1f8f28 R12: 0000000000040dc0
 R13: 000000050695a900 R14: 0000000000000017 R15: 0000000000000001
 FS:  00007f521f695740(0000) GS:ffff88806ce00000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00007f31b013f000 CR3: 000000006637e001 CR4: 0000000000370eb0
 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
 Call Trace:
  ? __kmalloc_track_caller+0x17a/0x570
  ? gcov_info_dup+0xfe/0x730
  ? gcov_event+0x88d/0xd30
  ? gcov_module_notifier+0xe9/0x100
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9
  ? mark_lock+0xba0/0xba0
  ? mark_lock+0xba0/0xba0
  ? notifier_call_chain+0xeb/0x170
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? do_syscall_64+0x45/0x70
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9
  ? warn_alloc+0x130/0x130
  ? lock_acquire+0x1f2/0xa30
  ? fs_reclaim_acquire+0x1f/0x70
  ? fs_reclaim_release+0x1f/0x50
  ? __kasan_check_read+0x1d/0x30
  ? reacquire_held_locks+0x420/0x420
  ? reacquire_held_locks+0x420/0x420
  kmalloc_order+0x3f/0xc0
  kmalloc_order_trace+0x24/0x220
  __kmalloc+0x41b/0x5a0
  ? gcov_info_dup+0xfe/0x730
  ? memcpy+0x73/0xa0
  gcov_info_dup+0x176/0x730
  gcov_event+0x88d/0xd30
  gcov_module_notifier+0xe9/0x100
  notifier_call_chain+0xeb/0x170
  blocking_notifier_call_chain+0x75/0xc0
  __x64_sys_delete_module+0x326/0x5a0
  ? do_init_module+0x810/0x810
  ? syscall_enter_from_user_mode+0x40/0x420
  ? trace_hardirqs_on+0x45/0xb0
  ? syscall_enter_from_user_mode+0x40/0x420
  do_syscall_64+0x45/0x70
  entry_SYSCALL_64_after_hwframe+0x44/0xa9
 RIP: 0033:0x7f521f7c531b
 Code: 73 01 c3 48 8b 0d 7d 0b 0c 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa b8 b0 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 4d 0b 0c 00 f7 d8 64 89 01 48
 RSP: 002b:00007ffe1bd4af48 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0
 RAX: ffffffffffffffda RBX: 0000561a3eae0910 RCX: 00007f521f7c531b
 RDX: 0000000000000000 RSI: 0000000000000800 RDI: 0000561a3eae0978
 RBP: 0000561a3eae0910 R08: 1999999999999999 R09: 0000000000000000
 R10: 00007f521f839ac0 R11: 0000000000000206 R12: 0000000000000000
 R13: 0000561a3eae0978 R14: 0000000000000000 R15: 0000561a3eae84d0
 irq event stamp: 326464
 hardirqs last  enabled at (326463): [<ffffffff832ecdde>] _raw_spin_unlock_irqrestore+0x8e/0xb0
 hardirqs last disabled at (326464): [<ffffffff832ec994>] _raw_spin_lock_irqsave+0x34/0xa0
 hardirqs last disabled at (326464): [<ffffffff832ec994>] _raw_spin_lock_irqsave+0x34/0xa0
 softirqs last  enabled at (320794): [<ffffffff83600931>] __do_softirq+0x931/0xbc4
 softirqs last disabled at (320789): [<ffffffff83400f2f>] asm_call_on_stack+0xf/0x20
 ---[ end trace 065ea9cc2ba144a6 ]---

[3] https://bugzilla.kernel.org/show_bug.cgi?id=208885#c1

Cc Colin Ian King <[email protected]>

Signed-off-by Leon Romanovsky <[email protected]>

---
I have a strong feeling that this solution is not correct, but don't know how to do it right. The problem exists and reproducable in seconds.

---
 kernel/gcov/gcc_4_7.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)

kernel/gcov/gcc_4_7.c +0 -1

index 908fdf5098c3..357ef839cdd3 100644
@@ -275,20 +275,23 @@ struct gcov_info *gcov_info_dup(struct gcov_info *info)
 	size_t fi_size; /* function info size */
 	size_t cv_size; /* counter values size */
 
-	dup = kmemdup(info, sizeof(*dup), GFP_KERNEL);

kernel/gcov/gcc_4_7.c struct gcov_info *gcov_info_dup(struct gcov_info *info) · line 284

Quoted 10 lines

kernel/gcov/gcc_4_7.c +4 -3

+	dup = kzalloc(sizeof(*dup), GFP_KERNEL);
 	if (!dup)
 		return NULL;
 
-	dup->next = NULL;
-	dup->filename = NULL;
-	dup->functions = NULL;
+	dup->version = info->version;
+	dup->stamp = info->stamp;
+	for (fi_idx = 0; i < GCOV_COUNTERS; i++)

This loop refers to i as a counter but it's not declared.

Quoted 28 lines

kernel/gcov/gcc_4_7.c +7 -5

+		dup->merge[i] = info->merge[i];
+	dup->n_functions = info->n_functions;
 
-	dup->filename = kstrdup(info->filename, GFP_KERNEL);
+	dup->filename = kstrdup_const(info->filename, GFP_KERNEL);
 	if (!dup->filename)
 		goto err_free;
 
-	dup->functions = kcalloc(info->n_functions,
-				 sizeof(struct gcov_fn_info *), GFP_KERNEL);
+	dup->functions =
+		kcalloc(info->n_functions, sizeof(struct gcov_fn_info *),
+			GFP_KERNEL | __GFP_NOWARN);
 	if (!dup->functions)
 		goto err_free;

@@ -359,7 +362,7 @@ void gcov_info_free(struct gcov_info *info)
 
 free_info:
 	kfree(info->functions);
-	kfree(info->filename);
+	kfree_const(info->filename);
 	kfree(info);
 }
 
--

2.26.2

Leon Romanovsky

author

2020-08-30 06:45 UTC #4 lore raw

Quoted 158 lines · Colin Ian King

Quoted 156 lines · Leon Romanovsky

From: Leon Romanovsky <[email protected]>

The kernel compiled with GCC 10.2.1 and KASAN together with GCOV enabled produces the following splats while reloading modules.

First splat [1] is generated due to the situation that gcov_info can be both user and kernel pointer, the memcpy() during kmemdup() causes to this. As a possible solution copy fields manually.

[1]
==================================================================

 BUG: KASAN: global-out-of-bounds in kmemdup+0x43/0x70
 Read of size 120 at addr ffffffffa0d2c780 by task modprobe/296

 CPU: 0 PID: 296 Comm: modprobe Not tainted 5.9.0-rc1+ #1860
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04 /01/2014
 Call Trace:
  ? dump_stack+0x128/0x1af
  ? print_address_description.constprop.0+0x2c/0x3f0
  ? _raw_spin_lock_irqsave+0x34/0xa0
  ? __kasan_check_read+0x1d/0x30
  ? kmemdup+0x43/0x70
  ? kmemdup+0x43/0x70
  ? gcov_info_dup+0x2d/0x730
  ? __kasan_check_write+0x20/0x30
  ? __mutex_unlock_slowpath+0x10d/0x740
  ? gcov_event+0x88d/0xd30
  ? gcov_module_notifier+0xe9/0x100
  ? notifier_call_chain+0xeb/0x170
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? do_init_module+0x810/0x810
  ? syscall_enter_from_user_mode+0x40/0x420
  ? trace_hardirqs_on+0x45/0xb0
  ? syscall_enter_from_user_mode+0x40/0x420
  ? do_syscall_64+0x45/0x70
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9

 The buggy address belongs to the variable:
  __gcov_.uverbs_attr_get_obj+0x60/0xfffffffffff778e0 [mlx5_ib]

 Memory state around the buggy address:
  ffffffffa0d2c680: 00 f9 f9 f9 f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9
  ffffffffa0d2c700: f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9 f9 f9 f9 f9
 >ffffffffa0d2c780: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 f9 f9
                                                              ^
  ffffffffa0d2c800: f9 f9 f9 f9 00 00 00 00 00 f9 f9 f9 f9 f9 f9 f9
  ffffffffa0d2c880: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 ==================================================================
 Disabling lock debugging due to kernel taint
 gcov: could not save data for '/home/leonro/src/kernel/drivers/infiniband/hw/mlx5/std_types.gcda' (out o

f memory)

[2]
Colin has similar error [3].

 ------------[ cut here ]------------
 WARNING: CPU: 0 PID: 296 at mm/page_alloc.c:4859 __alloc_pages_nodemask+0x670/0x3190
 Modules linked in: mlx5_ib(-) mlx5_core mlxfw ptp ib_ipoib pps_core rdma_ucm rdma_cm iw_cm ib_cm ib_umad  ib_uverbs ib_core
 CPU: 0 PID: 296 Comm: modprobe Tainted: G    B             5.9.0-rc1+ #1860
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04 /01/2014
 RIP: 0010:__alloc_pages_nodemask+0x670/0x3190
 Code: e9 af fc ff ff 48 83 05 fd 28 90 05 01 81 e7 00 20 00 00 48 c7 44 24 28 00 00 00 00 0f 85 fb fd ff  ff 48 83 05 f0 28 90 05 01 <0f> 0b 48 83 05 ee 28 90 05 01 48 83 05 ee 28 90 05 01 e9 dc fd ff
 RSP: 0018:ffff88805f7ffa28 EFLAGS: 00010202
 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 1ffff1100befff5e
 RDX: 0000000000000000 RSI: 0000000000000017 RDI: 0000000000000000
 RBP: 000000050695a900 R08: ffff888060fc7900 R09: ffff888060fc793b
 R10: ffffed100c1f8f27 R11: ffffed100c1f8f28 R12: 0000000000040dc0
 R13: 000000050695a900 R14: 0000000000000017 R15: 0000000000000001
 FS:  00007f521f695740(0000) GS:ffff88806ce00000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00007f31b013f000 CR3: 000000006637e001 CR4: 0000000000370eb0
 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
 Call Trace:
  ? __kmalloc_track_caller+0x17a/0x570
  ? gcov_info_dup+0xfe/0x730
  ? gcov_event+0x88d/0xd30
  ? gcov_module_notifier+0xe9/0x100
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9
  ? mark_lock+0xba0/0xba0
  ? mark_lock+0xba0/0xba0
  ? notifier_call_chain+0xeb/0x170
  ? blocking_notifier_call_chain+0x75/0xc0
  ? __x64_sys_delete_module+0x326/0x5a0
  ? do_syscall_64+0x45/0x70
  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9
  ? warn_alloc+0x130/0x130
  ? lock_acquire+0x1f2/0xa30
  ? fs_reclaim_acquire+0x1f/0x70
  ? fs_reclaim_release+0x1f/0x50
  ? __kasan_check_read+0x1d/0x30
  ? reacquire_held_locks+0x420/0x420
  ? reacquire_held_locks+0x420/0x420
  kmalloc_order+0x3f/0xc0
  kmalloc_order_trace+0x24/0x220
  __kmalloc+0x41b/0x5a0
  ? gcov_info_dup+0xfe/0x730
  ? memcpy+0x73/0xa0
  gcov_info_dup+0x176/0x730
  gcov_event+0x88d/0xd30
  gcov_module_notifier+0xe9/0x100
  notifier_call_chain+0xeb/0x170
  blocking_notifier_call_chain+0x75/0xc0
  __x64_sys_delete_module+0x326/0x5a0
  ? do_init_module+0x810/0x810
  ? syscall_enter_from_user_mode+0x40/0x420
  ? trace_hardirqs_on+0x45/0xb0
  ? syscall_enter_from_user_mode+0x40/0x420
  do_syscall_64+0x45/0x70
  entry_SYSCALL_64_after_hwframe+0x44/0xa9
 RIP: 0033:0x7f521f7c531b
 Code: 73 01 c3 48 8b 0d 7d 0b 0c 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa b8 b0 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 4d 0b 0c 00 f7 d8 64 89 01 48
 RSP: 002b:00007ffe1bd4af48 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0
 RAX: ffffffffffffffda RBX: 0000561a3eae0910 RCX: 00007f521f7c531b
 RDX: 0000000000000000 RSI: 0000000000000800 RDI: 0000561a3eae0978
 RBP: 0000561a3eae0910 R08: 1999999999999999 R09: 0000000000000000
 R10: 00007f521f839ac0 R11: 0000000000000206 R12: 0000000000000000
 R13: 0000561a3eae0978 R14: 0000000000000000 R15: 0000561a3eae84d0
 irq event stamp: 326464
 hardirqs last  enabled at (326463): [<ffffffff832ecdde>] _raw_spin_unlock_irqrestore+0x8e/0xb0
 hardirqs last disabled at (326464): [<ffffffff832ec994>] _raw_spin_lock_irqsave+0x34/0xa0
 hardirqs last disabled at (326464): [<ffffffff832ec994>] _raw_spin_lock_irqsave+0x34/0xa0
 softirqs last  enabled at (320794): [<ffffffff83600931>] __do_softirq+0x931/0xbc4
 softirqs last disabled at (320789): [<ffffffff83400f2f>] asm_call_on_stack+0xf/0x20
 ---[ end trace 065ea9cc2ba144a6 ]---

[3] https://bugzilla.kernel.org/show_bug.cgi?id=208885#c1

Cc Colin Ian King <[email protected]>

Signed-off-by Leon Romanovsky <[email protected]>

---
I have a strong feeling that this solution is not correct, but don't know how to do it right. The problem exists and reproducable in seconds.

---
 kernel/gcov/gcc_4_7.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)

kernel/gcov/gcc_4_7.c +1 -1

index 908fdf5098c3..357ef839cdd3 100644
@@ -275,20 +275,23 @@ struct gcov_info *gcov_info_dup(struct gcov_info *info)
      size_t fi_size; /* function info size */
      size_t cv_size; /* counter values size */
 
-     dup = kmemdup(info, sizeof(*dup), GFP_KERNEL);
+     dup = kzalloc(sizeof(*dup), GFP_KERNEL);
      if (!dup)
              return NULL;

Quoted 8 lines

kernel/gcov/gcc_4_7.c struct gcov_info *gcov_info_dup(struct gcov_info *info) · line 284
Quoted 6 lines
kernel/gcov/gcc_4_7.c +3 -3
-     dup->next = NULL;
-     dup->filename = NULL;
-     dup->functions = NULL;
+     dup->version = info->version;
+     dup->stamp = info->stamp;
+     for (fi_idx = 0; i < GCOV_COUNTERS; i++)
This loop refers to i as a counter but it's not declared.

Thanks, I wrote it above.
https://lore.kernel.org/lkml/20200827151333.GB2909436@unreal

The thing is that I would expect to hear any feedback if this GCC bug or not and how to solve it.
We tried to GCC 9 and gcov worked there.

Thanks