Skip to contentSkip to conversation

uprobes: Fix race in uprobe_free_utask

PATCH Applied Jiri Olsa 6 participants 6 messages Jan 9, 2025 – Jan 10, 2025

Subsystems (5)
x86 BPF perf events Tracing +1 more subsystems
lore
Share & tools
lore

Jiri Olsa

author reviewer
#0 lore raw

PATCH 1/1 uprobes: Fix race in uprobe_free_utask

Max Makarov reported kernel panic [1] in perf user callchain code.

The reason for that is the race between uprobe_free_utask and bpf profiler code doing the perf user stack unwind and is triggered within uprobe_free_utask function:
- after current->utask is freed and
- before current->utask is set to NULL

 general protection fault, probably for non-canonical address 0x9e759c37ee555c76: 0000 [#1] SMP PTI
 RIP: 0010:is_uprobe_at_func_entry+0x28/0x80
 ...
  ? die_addr+0x36/0x90
  ? exc_general_protection+0x217/0x420
  ? asm_exc_general_protection+0x26/0x30
  ? is_uprobe_at_func_entry+0x28/0x80
  perf_callchain_user+0x20a/0x360
  get_perf_callchain+0x147/0x1d0
  bpf_get_stackid+0x60/0x90
  bpf_prog_9aac297fb833e2f5_do_perf_event+0x434/0x53b
  ? __smp_call_single_queue+0xad/0x120
  bpf_overflow_handler+0x75/0x110
  ...
  asm_sysvec_apic_timer_interrupt+0x1a/0x20
 RIP: 0010:__kmem_cache_free+0x1cb/0x350
 ...
  ? uprobe_free_utask+0x62/0x80
  ? acct_collect+0x4c/0x220
  uprobe_free_utask+0x62/0x80
  mm_release+0x12/0xb0
  do_exit+0x26b/0xaa0
  __x64_sys_exit+0x1b/0x20
  do_syscall_64+0x5a/0x80

It can be easily reproduced by running following commands in separate terminals:

  # while :; do bpftrace -e 'uprobe:/bin/ls:_start  { printf("hit\n"); }' -c ls; done
  # bpftrace -e 'profile:hz:100000 { @[ustack()] = count(); }'

Fixing this by making sure current->utask pointer is set to NULL before we start to release the utask object.

[1] https://github.com/grafana/pyroscope/issues/3673

---
 kernel/events/uprobes.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/events/uprobes.c +1 -1
index fa04b14a7d72..5d71ef85420c 100644
@@ -1915,6 +1915,7 @@ void uprobe_free_utask(struct task_struct *t)
 	if (!utask)
 		return;
 
+	t->utask = NULL;
 	WARN_ON_ONCE(utask->active_uprobe || utask->xol_vaddr);
 
 	timer_delete_sync(&utask->ri_timer);
@@ -1924,7 +1925,6 @@ void uprobe_free_utask(struct task_struct *t)
 		ri = free_ret_instance(ri, true /* cleanup_hprobe */);
 
 	kfree(utask);
-	t->utask = NULL;
 }
 
 #define RI_TIMER_PERIOD (HZ / 10) /* 100 ms */
Signature · 1 line
2.47.0
maintainer Acked-by on v1 1/1
#1 lore raw
Quoted 12 lines · Jiri Olsa
kernel/events/uprobes.c +1 -0
@@ -1915,6 +1915,7 @@ void uprobe_free_utask(struct task_struct *t)
 	if (!utask)
 		return;
 
+	t->utask = NULL;
 	WARN_ON_ONCE(utask->active_uprobe || utask->xol_vaddr);
 
 	timer_delete_sync(&utask->ri_timer);

kernel/events/uprobes.c void uprobe_free_utask(struct task_struct *t) · line 1927

Quoted 4 lines
kernel/events/uprobes.c +0 -1
@@ -1924,7 +1925,6 @@ void uprobe_free_utask(struct task_struct *t)
 		ri = free_ret_instance(ri, true /* cleanup_hprobe */);
 
 	kfree(utask);
-	t->utask = NULL;
maintainer
#2 lore raw
In reply to Jiri Olsa
Quoted 33 lines · Jiri Olsa

Max Makarov reported kernel panic [1] in perf user callchain code.

The reason for that is the race between uprobe_free_utask and bpf profiler code doing the perf user stack unwind and is triggered within uprobe_free_utask function:
- after current->utask is freed and
- before current->utask is set to NULL

  general protection fault, probably for non-canonical address 0x9e759c37ee555c76: 0000 [#1] SMP PTI
  RIP: 0010:is_uprobe_at_func_entry+0x28/0x80
  ...
   ? die_addr+0x36/0x90
   ? exc_general_protection+0x217/0x420
   ? asm_exc_general_protection+0x26/0x30
   ? is_uprobe_at_func_entry+0x28/0x80
   perf_callchain_user+0x20a/0x360
   get_perf_callchain+0x147/0x1d0
   bpf_get_stackid+0x60/0x90
   bpf_prog_9aac297fb833e2f5_do_perf_event+0x434/0x53b
   ? __smp_call_single_queue+0xad/0x120
   bpf_overflow_handler+0x75/0x110
   ...
   asm_sysvec_apic_timer_interrupt+0x1a/0x20
  RIP: 0010:__kmem_cache_free+0x1cb/0x350
  ...
   ? uprobe_free_utask+0x62/0x80
   ? acct_collect+0x4c/0x220
   uprobe_free_utask+0x62/0x80
   mm_release+0x12/0xb0
   do_exit+0x26b/0xaa0
   __x64_sys_exit+0x1b/0x20
   do_syscall_64+0x5a/0x80
Quoted 8 lines

It can be easily reproduced by running following commands in separate terminals:

   # while :; do bpftrace -e 'uprobe:/bin/ls:_start  { printf("hit\n"); }' -c ls; done
   # bpftrace -e 'profile:hz:100000 { @[ustack()] = count(); }'

Fixing this by making sure current->utask pointer is set to NULL before we start to release the utask object.

Lets add Fixes tag for stable team:

Fixes cfa7f3d2c526 ("perf,x86: avoid missing caller address in stack traces captured in uprobe")
Quoted 3 lines

[1] https://github.com/grafana/pyroscope/issues/3673

fwiw, the other version we were potentially thinking of was below, but just moving the t->utask NULL assignment seemed better.

Thanks,
Daniel

arch/x86/events/core.c +2 -0
index c75c482d4c52..05f9cedf2691 100644
@@ -2835,6 +2835,8 @@ static bool is_uprobe_at_func_entry(struct pt_regs *regs)
 
        if (!current->utask)
                return false;
+       if (!current->utask->active_uprobe)
+               return false;
 
        auprobe = current->utask->auprobe;
        if (!auprobe)
#3 lore raw
Quoted 40 lines · Daniel Borkmann
Quoted 39 lines · Jiri Olsa

Max Makarov reported kernel panic [1] in perf user callchain code.

The reason for that is the race between uprobe_free_utask and bpf profiler code doing the perf user stack unwind and is triggered within uprobe_free_utask function:
- after current->utask is freed and
- before current->utask is set to NULL

  general protection fault, probably for non-canonical address 0x9e759c37ee555c76: 0000 [#1] SMP PTI
  RIP: 0010:is_uprobe_at_func_entry+0x28/0x80
  ...
   ? die_addr+0x36/0x90
   ? exc_general_protection+0x217/0x420
   ? asm_exc_general_protection+0x26/0x30
   ? is_uprobe_at_func_entry+0x28/0x80
   perf_callchain_user+0x20a/0x360
   get_perf_callchain+0x147/0x1d0
   bpf_get_stackid+0x60/0x90
   bpf_prog_9aac297fb833e2f5_do_perf_event+0x434/0x53b
   ? __smp_call_single_queue+0xad/0x120
   bpf_overflow_handler+0x75/0x110
   ...
   asm_sysvec_apic_timer_interrupt+0x1a/0x20
  RIP: 0010:__kmem_cache_free+0x1cb/0x350
  ...
   ? uprobe_free_utask+0x62/0x80
   ? acct_collect+0x4c/0x220
   uprobe_free_utask+0x62/0x80
   mm_release+0x12/0xb0
   do_exit+0x26b/0xaa0
   __x64_sys_exit+0x1b/0x20
   do_syscall_64+0x5a/0x80

It can be easily reproduced by running following commands in separate terminals:

   # while :; do bpftrace -e 'uprobe:/bin/ls:_start  { printf("hit\n"); }' -c ls; done
   # bpftrace -e 'profile:hz:100000 { @[ustack()] = count(); }'
Quoted 6 lines
Quoted 2 lines

Fixing this by making sure current->utask pointer is set to NULL before we start to release the utask object.

Lets add Fixes tag for stable team:

Fixes cfa7f3d2c526 ("perf,x86: avoid missing caller address in stack traces captured in uprobe")

ugh right, thanks for finding that

jirka

Quoted 24 lines
Quoted 4 lines

[1] https://github.com/grafana/pyroscope/issues/3673

fwiw, the other version we were potentially thinking of was below, but just moving the t->utask NULL assignment seemed better.

Thanks,
Daniel

arch/x86/events/core.c +2 -0
index c75c482d4c52..05f9cedf2691 100644
@@ -2835,6 +2835,8 @@ static bool is_uprobe_at_func_entry(struct pt_regs *regs)
 
        if (!current->utask)
                return false;
+       if (!current->utask->active_uprobe)
+               return false;
 
        auprobe = current->utask->auprobe;
        if (!auprobe)
Acked-by on v1 1/1
#4 lore raw
In reply to Jiri Olsa
Quoted 43 lines · Jiri Olsa

Max Makarov reported kernel panic [1] in perf user callchain code.

The reason for that is the race between uprobe_free_utask and bpf profiler code doing the perf user stack unwind and is triggered within uprobe_free_utask function:
- after current->utask is freed and
- before current->utask is set to NULL

 general protection fault, probably for non-canonical address 0x9e759c37ee555c76: 0000 [#1] SMP PTI
 RIP: 0010:is_uprobe_at_func_entry+0x28/0x80
 ...
  ? die_addr+0x36/0x90
  ? exc_general_protection+0x217/0x420
  ? asm_exc_general_protection+0x26/0x30
  ? is_uprobe_at_func_entry+0x28/0x80
  perf_callchain_user+0x20a/0x360
  get_perf_callchain+0x147/0x1d0
  bpf_get_stackid+0x60/0x90
  bpf_prog_9aac297fb833e2f5_do_perf_event+0x434/0x53b
  ? __smp_call_single_queue+0xad/0x120
  bpf_overflow_handler+0x75/0x110
  ...
  asm_sysvec_apic_timer_interrupt+0x1a/0x20
 RIP: 0010:__kmem_cache_free+0x1cb/0x350
 ...
  ? uprobe_free_utask+0x62/0x80
  ? acct_collect+0x4c/0x220
  uprobe_free_utask+0x62/0x80
  mm_release+0x12/0xb0
  do_exit+0x26b/0xaa0
  __x64_sys_exit+0x1b/0x20
  do_syscall_64+0x5a/0x80

It can be easily reproduced by running following commands in separate terminals:

  # while :; do bpftrace -e 'uprobe:/bin/ls:_start  { printf("hit\n"); }' -c ls; done
  # bpftrace -e 'profile:hz:100000 { @[ustack()] = count(); }'

Fixing this by making sure current->utask pointer is set to NULL before we start to release the utask object.

Quoted 7 lines

[1] https://github.com/grafana/pyroscope/issues/3673

---
 kernel/events/uprobes.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

ah, it's interrupt/NMI checking current->utask, makes total sense, thanks for the fix!

Quoted 23 lines
kernel/events/uprobes.c +1 -2
index fa04b14a7d72..5d71ef85420c 100644
@@ -1915,6 +1915,7 @@ void uprobe_free_utask(struct task_struct *t)
        if (!utask)
                return;
 
+       t->utask = NULL;
        WARN_ON_ONCE(utask->active_uprobe || utask->xol_vaddr);
 
        timer_delete_sync(&utask->ri_timer);
@@ -1924,7 +1925,6 @@ void uprobe_free_utask(struct task_struct *t)
                ri = free_ret_instance(ri, true /* cleanup_hprobe */);
 
        kfree(utask);
-       t->utask = NULL;
 }
 
 #define RI_TIMER_PERIOD (HZ / 10) /* 100 ms */
--

2.47.0

merge bot Applied
#5 lore raw
In reply to Jiri Olsa
The following commit has been merged into the perf/urgent branch of tip:

The following commit has been merged into the perf/urgent branch of tip:

Commit-ID:     b583ef82b671c9a752fbe3e95bd4c1c51eab764d
Gitweb:        https://git.kernel.org/tip/b583ef82b671c9a752fbe3e95bd4c1c51eab764d
Author:        Jiri Olsa <[email protected]>
AuthorDate:    Thu, 09 Jan 2025 15:14:40 +01:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Fri, 10 Jan 2025 09:28:01 +01:00

uprobes: Fix race in uprobe_free_utask

Max Makarov reported kernel panic [1] in perf user callchain code.

The reason for that is the race between uprobe_free_utask and bpf profiler code doing the perf user stack unwind and is triggered within uprobe_free_utask function:
- after current->utask is freed and
- before current->utask is set to NULL

 general protection fault, probably for non-canonical address 0x9e759c37ee555c76: 0000 [#1] SMP PTI
 RIP: 0010:is_uprobe_at_func_entry+0x28/0x80
 ...
  ? die_addr+0x36/0x90
  ? exc_general_protection+0x217/0x420
  ? asm_exc_general_protection+0x26/0x30
  ? is_uprobe_at_func_entry+0x28/0x80
  perf_callchain_user+0x20a/0x360
  get_perf_callchain+0x147/0x1d0
  bpf_get_stackid+0x60/0x90
  bpf_prog_9aac297fb833e2f5_do_perf_event+0x434/0x53b
  ? __smp_call_single_queue+0xad/0x120
  bpf_overflow_handler+0x75/0x110
  ...
  asm_sysvec_apic_timer_interrupt+0x1a/0x20
 RIP: 0010:__kmem_cache_free+0x1cb/0x350
 ...
  ? uprobe_free_utask+0x62/0x80
  ? acct_collect+0x4c/0x220
  uprobe_free_utask+0x62/0x80
  mm_release+0x12/0xb0
  do_exit+0x26b/0xaa0
  __x64_sys_exit+0x1b/0x20
  do_syscall_64+0x5a/0x80

It can be easily reproduced by running following commands in separate terminals:

  # while :; do bpftrace -e 'uprobe:/bin/ls:_start  { printf("hit\n"); }' -c ls; done
  # bpftrace -e 'profile:hz:100000 { @[ustack()] = count(); }'

Fixing this by making sure current->utask pointer is set to NULL before we start to release the utask object.

[1] https://github.com/grafana/pyroscope/issues/3673

Fixes cfa7f3d2c526 ("perf,x86: avoid missing caller address in stack traces captured in uprobe")
---
 kernel/events/uprobes.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/events/uprobes.c +1 -1
index fa04b14..5d71ef8 100644
@@ -1915,6 +1915,7 @@ void uprobe_free_utask(struct task_struct *t)
 	if (!utask)
 		return;
 
+	t->utask = NULL;
 	WARN_ON_ONCE(utask->active_uprobe || utask->xol_vaddr);
 
 	timer_delete_sync(&utask->ri_timer);
@@ -1924,7 +1925,6 @@ void uprobe_free_utask(struct task_struct *t)
 		ri = free_ret_instance(ri, true /* cleanup_hprobe */);
 
 	kfree(utask);
-	t->utask = NULL;
 }
 
 #define RI_TIMER_PERIOD (HZ / 10) /* 100 ms */