serial: core: fix infinite loop in handle_tx() for PORT_UNKNOWN

PATCH In review Jiayuan Chen 2 participants 5 messages Feb 4 TTY layer and serial drivers

lore

Share & tools

lore

Jiayuan Chen

author

2026-02-04 07:43 UTC #0 lore raw

PATCH 1/1 serial: core: fix infinite loop in handle_tx() for PORT_UNKNOWN

From: Jiayuan Chen <[email protected]>

uart_write_room() and uart_write() behave inconsistently when xmit_buf is NULL (which happens for PORT_UNKNOWN ports that were never properly initialized):

- uart_write_room() returns kfifo_avail() which can be > 0
- uart_write() checks xmit_buf and returns 0 if NULL

This inconsistency causes an infinite loop in drivers that rely on tty_write_room() to determine if they can write:

  while (tty_write_room(tty) > 0) {
      written = tty->ops->write(...);
      // written is always 0, loop never exits
  }

For example, caif_serial's handle_tx() enters an infinite loop when used with PORT_UNKNOWN serial ports, causing system hangs.

Fix by making uart_write_room() also check xmit_buf and return 0 if it's NULL, consistent with uart_write().

Reproducer: https://gist.github.com/mrpre/d9a694cc0e19828ee3bc3b37983fde13

Fixes 9b27105b4a44 ("net-caif-driver: add CAIF serial driver (ldisc)")

Signed-off-by Jiayuan Chen <[email protected]>

---
 drivers/tty/serial/serial_core.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

drivers/tty/serial/serial_core.c +4 -1

index 2805cad10511..0b2edf185cc7 100644
@@ -643,7 +643,10 @@ static unsigned int uart_write_room(struct tty_struct *tty)
 	unsigned int ret;
 
 	port = uart_port_ref_lock(state, &flags);
-	ret = kfifo_avail(&state->port.xmit_fifo);
+	if (!state->port.xmit_buf)
+		ret = 0;
+	else
+		ret = kfifo_avail(&state->port.xmit_fifo);
 	uart_port_unlock_deref(port, flags);
 	return ret;
 }

Signature · 1 line

2.43.0

Greg Kroah-Hartman

maintainer

2026-02-04 08:20 UTC #1 lore raw

Quoted 5 lines · Jiayuan Chen

From: Jiayuan Chen <[email protected]>
uart_write_room() and uart_write() behave inconsistently when xmit_buf is NULL (which happens for PORT_UNKNOWN ports that were never properly initialized):

How does this happen? Why were they not initialized properly, what drivers/hardware cause this?

Quoted 14 lines

- uart_write_room() returns kfifo_avail() which can be > 0
- uart_write() checks xmit_buf and returns 0 if NULL
This inconsistency causes an infinite loop in drivers that rely on tty_write_room() to determine if they can write:
  while (tty_write_room(tty) > 0) {
      written = tty->ops->write(...);
      // written is always 0, loop never exits
  }
For example, caif_serial's handle_tx() enters an infinite loop when used with PORT_UNKNOWN serial ports, causing system hangs.

Quoted 6 lines

Fix by making uart_write_room() also check xmit_buf and return 0 if it's NULL, consistent with uart_write().
Reproducer: https://gist.github.com/mrpre/d9a694cc0e19828ee3bc3b37983fde13
Fixes 9b27105b4a44 ("net-caif-driver: add CAIF serial driver (ldisc)")

This really isn't a fix for that driver, but rather something else.

Quoted 2 lines

Signed-off-by Jiayuan Chen <[email protected]>
Signed-off-by Jiayuan Chen <[email protected]>

This doesn't make sense, signing off twice for the same person?

As you did this from your shopee.com account, that should be sufficient.

Quoted 9 lines

---
 drivers/tty/serial/serial_core.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

drivers/tty/serial/serial_core.c +0 -0

index 2805cad10511..0b2edf185cc7 100644

drivers/tty/serial/serial_core.c static unsigned int uart_write_room(struct tty_struct *tty) · line 646

Quoted 5 lines

drivers/tty/serial/serial_core.c +1 -1

@@ -643,7 +643,10 @@ static unsigned int uart_write_room(struct tty_struct *tty)
 	unsigned int ret;
 
 	port = uart_port_ref_lock(state, &flags);
-	ret = kfifo_avail(&state->port.xmit_fifo);
+	if (!state->port.xmit_buf)

This feels odd. What ports have no transmit buffers? And why would this be the only check that is needed for such broken devices?

Maybe let's fix the root cause here, the driver that does not have a transmit buffer at all?

thanks,

greg k-h

Jiayuan Chen

author

2026-02-04 08:29 UTC #2 lore raw

Quoted 4 lines · Greg Kroah-Hartman

Quoted 1 line · Jiayuan Chen

Quoted 8 lines

Quoted 6 lines
From: Jiayuan Chen <[email protected]>
uart_write_room() and uart_write() behave inconsistently when xmit_buf is NULL (which happens for PORT_UNKNOWN ports that were never properly initialized):
How does this happen? Why were they not initialized properly, what drivers/hardware cause this?

In QEMU environment, /dev/ttyS3 is PORT_UNKNOWN type (no real UART hardware). When uart_port_startup() sees uport->type == PORT_UNKNOWN, it returns early without allocating xmit_buf:
if (uport->type == PORT_UNKNOWN)
return 1; // xmit_buf never allocated
So xmit_buf remains NULL.

Quoted 15 lines

Quoted 15 lines
- uart_write_room() returns kfifo_avail() which can be > 0
- uart_write() checks xmit_buf and returns 0 if NULL
This inconsistency causes an infinite loop in drivers that rely on tty_write_room() to determine if they can write:
while (tty_write_room(tty) > 0) {
written = tty->ops->write(...);
// written is always 0, loop never exits
}
For example, caif_serial's handle_tx() enters an infinite loop when used with PORT_UNKNOWN serial ports, causing system hangs.

Quoted 8 lines

Quoted 7 lines
Fix by making uart_write_room() also check xmit_buf and return 0 if it's NULL, consistent with uart_write().
Reproducer: https://gist.github.com/mrpre/d9a694cc0e19828ee3bc3b37983fde13
Fixes 9b27105b4a44 ("net-caif-driver: add CAIF serial driver (ldisc)")
This really isn't a fix for that driver, but rather something else.

You're right, this is awkward. The API inconsistency between uart_write_room() and uart_write() has existed since 2.6.12, but it only became visible as a deadloop when CAIF was introduced - because CAIF's handle_tx() relies on tty_write_room() to decide whether to call write().
The fix location is in uart, but the trigger condition requires CAIF (or similar drivers). I can remove the Fixes tag if you prefer.

Quoted 7 lines

Quoted 4 lines
Signed-off-by Jiayuan Chen <[email protected]>
Signed-off-by Jiayuan Chen <[email protected]>
This doesn't make sense, signing off twice for the same person?
As you did this from your shopee.com account, that should be sufficient.

Sorry, I'll keep only one Signed-off-by.

Quoted 16 lines

Quoted 12 lines
---
 drivers/tty/serial/serial_core.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c index 2805cad10511..0b2edf185cc7 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -643,7 +643,10 @@ static unsigned int uart_write_room(struct tty_struct *tty) unsigned int ret;
Quoted 4 lines
port = uart_port_ref_lock(state, &flags);
- ret = kfifo_avail(&state->port.xmit_fifo);
+ if (!state->port.xmit_buf)

Quoted 5 lines

This feels odd. What ports have no transmit buffers? And why would this be the only check that is needed for such broken devices?
Maybe let's fix the root cause here, the driver that does not have a transmit buffer at all?

Do you suggest we should prevent setting line discipline (like N_CAIF) on PORT_UNKNOWN ports? Or should CAIF check the port type before using it? Note that CAIF is currently in orphan status (no active maintainer), so I'm not sure about the process for modifying it. The serial core fix might be more straightforward.

I'm happy to take either direction.
Thanks,
Jiayuan

Quoted 4 lines

thanks,
greg k-h

Greg Kroah-Hartman

maintainer

2026-02-04 08:53 UTC #3 lore raw

Quoted 17 lines · Jiayuan Chen

Quoted 4 lines · Greg Kroah-Hartman
Quoted 1 line · Jiayuan Chen
Quoted 8 lines
Quoted 6 lines
From: Jiayuan Chen <[email protected]>
uart_write_room() and uart_write() behave inconsistently when xmit_buf is NULL (which happens for PORT_UNKNOWN ports that were never properly initialized):
How does this happen? Why were they not initialized properly, what drivers/hardware cause this?

Quoted 6 lines

In QEMU environment, /dev/ttyS3 is PORT_UNKNOWN type (no real UART hardware). When uart_port_startup() sees uport->type == PORT_UNKNOWN, it returns early without allocating xmit_buf:
if (uport->type == PORT_UNKNOWN)
return 1; // xmit_buf never allocated
So xmit_buf remains NULL.

But the flags for the port will have TTY_IO_ERROR set on it, which should hopefully mean that no data is attempted to be sent through this (or a ldisc would be bound to it.)

How does this port work at all? Why is QEMU advertising a broken port that can not do anything?

And is this the only place such a check would ever be needed? What changed recently to suddenly require this?

Quoted 22 lines

Quoted 22 lines
Quoted 22 lines
- uart_write_room() returns kfifo_avail() which can be > 0
- uart_write() checks xmit_buf and returns 0 if NULL
This inconsistency causes an infinite loop in drivers that rely on tty_write_room() to determine if they can write:
while (tty_write_room(tty) > 0) {
written = tty->ops->write(...);
// written is always 0, loop never exits
}
For example, caif_serial's handle_tx() enters an infinite loop when used with PORT_UNKNOWN serial ports, causing system hangs.
Fix by making uart_write_room() also check xmit_buf and return 0 if it's NULL, consistent with uart_write().
Reproducer: https://gist.github.com/mrpre/d9a694cc0e19828ee3bc3b37983fde13
Fixes 9b27105b4a44 ("net-caif-driver: add CAIF serial driver (ldisc)")

Quoted 8 lines

Quoted 1 line
This really isn't a fix for that driver, but rather something else.
You're right, this is awkward. The API inconsistency between uart_write_room() and uart_write() has existed since 2.6.12, but it only became visible as a deadloop when CAIF was introduced - because CAIF's handle_tx() relies on tty_write_room() to decide whether to call write().
The fix location is in uart, but the trigger condition requires CAIF (or similar drivers). I can remove the Fixes tag if you prefer.

Ok, I think this goes a bit deeper. This might be due to the kfifo rewrite of the serial drivers, as in older kernels we did not have a kfifo, so if it was not initialized the code checking path is much different.

As a "check" can you see if this fails for you on the latest 5.10.y tree? That is before the kfifo code was added to the uart layer.

Quoted 22 lines

Quoted 15 lines
Quoted 11 lines
---
 drivers/tty/serial/serial_core.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c index 2805cad10511..0b2edf185cc7 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -643,7 +643,10 @@ static unsigned int uart_write_room(struct tty_struct *tty) unsigned int ret;
Quoted 4 lines
port = uart_port_ref_lock(state, &flags);
- ret = kfifo_avail(&state->port.xmit_fifo);
+ if (!state->port.xmit_buf)
Quoted 5 lines
This feels odd. What ports have no transmit buffers? And why would this be the only check that is needed for such broken devices?
Maybe let's fix the root cause here, the driver that does not have a transmit buffer at all?

Quoted 5 lines

Do you suggest we should prevent setting line discipline (like N_CAIF) on PORT_UNKNOWN ports? Or should CAIF check the port type before using it? Note that CAIF is currently in orphan status (no active maintainer), so I'm not sure about the process for modifying it. The serial core fix might be more straightforward.

I think you found a real bug here, that is independent of the caif code, and might just be due to the kfifo stuff. See above for my questions here, and if so, your patch is correct, it's just that the Fixes: tag is a bit off.

thanks,

greg k-h

Jiayuan Chen

author

2026-02-04 09:57 UTC #4 lore raw

Quoted 32 lines · Greg Kroah-Hartman

Quoted 18 lines · Jiayuan Chen
2026/2/4 16:20, "Greg Kroah-Hartman" <[email protected] mailto:[email protected]?to=%22Greg%20Kroah-Hartman%22%20%3Cgregkh%40linuxfoundation.org%3E > wrote:
Quoted 7 lines · Jiayuan Chen
From: Jiayuan Chen <[email protected]>
uart_write_room() and uart_write() behave inconsistently when xmit_buf is NULL (which happens for PORT_UNKNOWN ports that were never properly initialized):
How does this happen? Why were they not initialized properly, what drivers/hardware cause this?
Quoted 7 lines
In QEMU environment, /dev/ttyS3 is PORT_UNKNOWN type (no real UART hardware). When uart_port_startup() sees uport->type == PORT_UNKNOWN, it returns early without allocating xmit_buf:
if (uport->type == PORT_UNKNOWN)
return 1; // xmit_buf never allocated
So xmit_buf remains NULL.
But the flags for the port will have TTY_IO_ERROR set on it, which should hopefully mean that no data is attempted to be sent through this (or a ldisc would be bound to it.)

Quoted 5 lines

How does this port work at all? Why is QEMU advertising a broken port that can not do anything?
And is this the only place such a check would ever be needed? What changed recently to suddenly require this?

This is an artificially constructed reproducer. I chose /dev/ttyS3 specifically because it's PORT_UNKNOWN in QEMU. In real-world usage, users wouldn't do this intentionally.

Quoted 32 lines

Quoted 23 lines
Quoted 16 lines
- uart_write_room() returns kfifo_avail() which can be > 0
- uart_write() checks xmit_buf and returns 0 if NULL
This inconsistency causes an infinite loop in drivers that rely on tty_write_room() to determine if they can write:
while (tty_write_room(tty) > 0) {
written = tty->ops->write(...);
// written is always 0, loop never exits
}
For example, caif_serial's handle_tx() enters an infinite loop when used with PORT_UNKNOWN serial ports, causing system hangs.
Quoted 7 lines
Fix by making uart_write_room() also check xmit_buf and return 0 if it's NULL, consistent with uart_write().
Reproducer: https://gist.github.com/mrpre/d9a694cc0e19828ee3bc3b37983fde13
Fixes 9b27105b4a44 ("net-caif-driver: add CAIF serial driver (ldisc)")
Quoted 9 lines
This really isn't a fix for that driver, but rather something else.
You're right, this is awkward. The API inconsistency between uart_write_room() and uart_write() has existed since 2.6.12, but it only became visible as a deadloop when CAIF was introduced - because CAIF's handle_tx() relies on tty_write_room() to decide whether to call write().
The fix location is in uart, but the trigger condition requires CAIF (or similar drivers). I can remove the Fixes tag if you prefer.

Quoted 7 lines

Ok, I think this goes a bit deeper. This might be due to the kfifo rewrite of the serial drivers, as in older kernels we did not have a kfifo, so if it was not initialized the code checking path is much different.
As a "check" can you see if this fails for you on the latest 5.10.y tree? That is before the kfifo code was added to the uart layer.

This issue still exists in 5.10.248

[   56.519143] watchdog: BUG: soft lockup - CPU#2 stuck for 22s! [caif_deadloop_r:457]
[   56.520868] Modules linked in:
[   56.520903] CPU: 2 PID: 457 Comm: caif_deadloop_r Not tainted 5.10.248 #1
[   56.520914] Hardware name: QEMU Ubuntu 24.04 PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[   56.520971] RIP: 0010:_raw_spin_unlock_irqrestore+0x15/0x20
[   56.520977] Code: e8 a0 5f 38 ff 4c 29 e8 49 39 c6 73 d8 80 0b 04 eb 8d cc cc cc 0f 1f 44 00 00 55 48 89 e5 e8 8a 4e 3b ff 66 90 48 89 f7 57 9d <0f> 1f 44 00 00 5d c3 cc cc cc cc 0f 1f 47
[   56.520986] RSP: 0018:ffffc90000f8bb60 EFLAGS: 00000282
[   56.520988] RAX: 0000000000000001 RBX: ffff888100b984e0 RCX: ffff8881024eb800
[   56.520990] RDX: 0000000000000001 RSI: 0000000000000282 RDI: 0000000000000282
[   56.520991] RBP: ffffc90000f8bb60 R08: ffff8881024eb800 R09: 0000000000000000
[   56.520992] R10: ffff88810086ed00 R11: 0000000000000000 R12: 0000000000000080
[   56.520993] R13: ffff888102423e10 R14: ffff8881024eb800 R15: ffffffff841eeb58
[   56.520996] FS:  00007f5c618c7740(0000) GS:ffff888137c00000(0000) knlGS:0000000000000000
[   56.520997] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   56.520998] CR2: 00007f1767cce200 CR3: 0000000008622005 CR4: 0000000000770ee0
[   56.521003] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[   56.521004] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[   56.521005] PKRU: 55555554
[   56.521010] Call Trace:
[   56.521087]  uart_write+0x1ec/0x240
[   56.521112]  handle_tx+0x9a/0x1a0
[   56.521115]  caif_xmit+0x61/0x70
[   56.521141]  dev_hard_start_xmit+0xa6/0x1e0
[   56.521144]  __dev_queue_xmit+0x7b3/0xaa0
[   56.521165]  ? packet_parse_headers+0x17a/0x250
[   56.521169]  dev_queue_xmit+0x10/0x20
[   56.521175]  packet_sendmsg+0x8eb/0x1740
[   56.521197]  ? __wake_up_common_lock+0x88/0xc0
[   56.521214]  __sock_sendmsg+0x70/0x80
[   56.521217]  __sys_sendto+0x142/0x190
[   56.521223]  __x64_sys_sendto+0x24/0x30
[   56.521233]  do_syscall_64+0x37/0x50
[   56.521236]  entry_SYSCALL_64_after_hwframe+0x67/0xd1
[   56.521251] RIP: 0033:0x7f5c619f60d7
[   56.521276] Code: c7 c0 ff ff ff ff eb be 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 80 3d 75 ef 0d 00 00 41 89 ca 74 10 b8 2c 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 69 c3 55 48 89 e5 50
[   56.521277] RSP: 002b:00007ffd7a4f64b8 EFLAGS: 00000202 ORIG_RAX: 000000000000002c
[   56.521279] RAX: ffffffffffffffda RBX: 00007ffd7a4f67a8 RCX: 00007f5c619f60d7
[   56.521281] RDX: 0000000000000080 RSI: 00007ffd7a4f64f0 RDI: 0000000000000004
[   56.521282] RBP: 00007ffd7a4f6680 R08: 00007ffd7a4f64d0 R09: 0000000000000014
[   56.521283] R10: 0000000000000000 R11: 0000000000000202 R12: 0000000000000001
[   56.521285] R13: 0000000000000000 R14: 000055c2c648ed58 R15: 00007f5c61b1a000

$ scripts/decode_stacktrace.sh vmlinux < dmesg.txt

[   56.519143] watchdog: BUG: soft lockup - CPU#2 stuck for 22s! [caif_deadloop_r:457]
[   56.520868] Modules linked in:
[   56.520903] CPU: 2 PID: 457 Comm: caif_deadloop_r Not tainted 5.10.248 #1
[   56.520914] Hardware name: QEMU Ubuntu 24.04 PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[   56.520971] RIP: 0010:_raw_spin_unlock_irqrestore (./arch/x86/include/asm/paravirt.h:653 ./include/linux/spinlock_api_smp.h:160 kernel/locking/spinlock.c:191)
[ 56.520977] Code: e8 a0 5f 38 ff 4c 29 e8 49 39 c6 73 d8 80 0b 04 eb 8d cc cc cc 0f 1f 44 00 00 55 48 89 e5 e8 8a 4e 3b ff 66 90 48 89 f7 57 9d <0f> 1f 44 00 00 5d c3 cc cc cc cc 0f 1f 47
All code
========
   0:	e8 a0 5f 38 ff       	call   0xffffffffff385fa5
   5:	4c 29 e8             	sub    %r13,%rax
   8:	49 39 c6             	cmp    %rax,%r14
   b:	73 d8                	jae    0xffffffffffffffe5
   d:	80 0b 04             	orb    $0x4,(%rbx)
  10:	eb 8d                	jmp    0xffffffffffffff9f
  12:	cc                   	int3
  13:	cc                   	int3
  14:	cc                   	int3
  15:	0f 1f 44 00 00       	nopl   0x0(%rax,%rax,1)
  1a:	55                   	push   %rbp
  1b:	48 89 e5             	mov    %rsp,%rbp
  1e:	e8 8a 4e 3b ff       	call   0xffffffffff3b4ead
  23:	66 90                	xchg   %ax,%ax
  25:	48 89 f7             	mov    %rsi,%rdi
  28:	57                   	push   %rdi
  29:	9d                   	popf
  2a:*	0f 1f 44 00 00       	nopl   0x0(%rax,%rax,1)		<-- trapping instruction
  2f:	5d                   	pop    %rbp
  30:	c3                   	ret
  31:	cc                   	int3
  32:	cc                   	int3
  33:	cc                   	int3
  34:	cc                   	int3
  35:	0f                   	.byte 0xf
  36:	1f                   	(bad)
  37:	47                   	rex.RXB

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

   0:	0f 1f 44 00 00       	nopl   0x0(%rax,%rax,1)
   5:	5d                   	pop    %rbp
   6:	c3                   	ret
   7:	cc                   	int3
   8:	cc                   	int3
   9:	cc                   	int3
   a:	cc                   	int3
   b:	0f                   	.byte 0xf
   c:	1f                   	(bad)
   d:	47                   	rex.RXB
[   56.520986] RSP: 0018:ffffc90000f8bb60 EFLAGS: 00000282
[   56.520988] RAX: 0000000000000001 RBX: ffff888100b984e0 RCX: ffff8881024eb800
[   56.520990] RDX: 0000000000000001 RSI: 0000000000000282 RDI: 0000000000000282
[   56.520991] RBP: ffffc90000f8bb60 R08: ffff8881024eb800 R09: 0000000000000000
[   56.520992] R10: ffff88810086ed00 R11: 0000000000000000 R12: 0000000000000080
[   56.520993] R13: ffff888102423e10 R14: ffff8881024eb800 R15: ffffffff841eeb58
[   56.520996] FS:  00007f5c618c7740(0000) GS:ffff888137c00000(0000) knlGS:0000000000000000
[   56.520997] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   56.520998] CR2: 00007f1767cce200 CR3: 0000000008622005 CR4: 0000000000770ee0
[   56.521003] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[   56.521004] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[   56.521005] PKRU: 55555554
[   56.521010] Call Trace:
[   56.521087] uart_write (drivers/tty/serial/serial_core.c:72 drivers/tty/serial/serial_core.c:598)
[   56.521112] handle_tx (drivers/net/caif/caif_serial.c:237)
[   56.521115] caif_xmit (drivers/net/caif/caif_serial.c:284)
[   56.521141] dev_hard_start_xmit (./include/linux/netdevice.h:4833 ./include/linux/netdevice.h:4847 net/core/dev.c:3601 net/core/dev.c:3617)
[   56.521144] __dev_queue_xmit (./include/linux/netdevice.h:3322 (discriminator 25) net/core/dev.c:4204 (discriminator 25))
[   56.521165] ? packet_parse_headers (./include/linux/skbuff.h:2616 (discriminator 1) net/packet/af_packet.c:1954 (discriminator 1))
[   56.521169] dev_queue_xmit (net/core/dev.c:4237)
[   56.521175] packet_sendmsg (net/packet/af_packet.c:3086 (discriminator 1) net/packet/af_packet.c:3118 (discriminator 1))
[   56.521197] ? __wake_up_common_lock (kernel/sched/wait.c:126 (discriminator 1))
[   56.521214] __sock_sendmsg (net/socket.c:651 (discriminator 1) net/socket.c:663 (discriminator 1))
[   56.521217] __sys_sendto (./include/linux/file.h:33 net/socket.c:2008)
[   56.521223] __x64_sys_sendto (net/socket.c:2013)
[   56.521233] do_syscall_64 (arch/x86/entry/common.c:46 (discriminator 1))
[   56.521236] entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:117)
[   56.521251] RIP: 0033:0x7f5c619f60d7
[ 56.521276] Code: c7 c0 ff ff ff ff eb be 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 80 3d 75 ef 0d 00 00 41 89 ca 74 10 b8 2c 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 69 c3 55 48 89 e5 50
All code
========
   0:	c7 c0 ff ff ff ff    	mov    $0xffffffff,%eax
   6:	eb be                	jmp    0xffffffffffffffc6
   8:	66 2e 0f 1f 84 00 00 	cs nopw 0x0(%rax,%rax,1)
   f:	00 00 00
  12:	90                   	nop
  13:	f3 0f 1e fa          	endbr64
  17:	80 3d 75 ef 0d 00 00 	cmpb   $0x0,0xdef75(%rip)        # 0xdef93
  1e:	41 89 ca             	mov    %ecx,%r10d
  21:	74 10                	je     0x33
  23:	b8 2c 00 00 00       	mov    $0x2c,%eax
  28:	0f 05                	syscall
  2a:*	48 3d 00 f0 ff ff    	cmp    $0xfffffffffffff000,%rax		<-- trapping instruction
  30:	77 69                	ja     0x9b
  32:	c3                   	ret
  33:	55                   	push   %rbp
  34:	48 89 e5             	mov    %rsp,%rbp
  37:	50                   	push   %rax

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

   0:	48 3d 00 f0 ff ff    	cmp    $0xfffffffffffff000,%rax
   6:	77 69                	ja     0x71
   8:	c3                   	ret
   9:	55                   	push   %rbp
   a:	48 89 e5             	mov    %rsp,%rbp
   d:	50                   	push   %rax
[   56.521277] RSP: 002b:00007ffd7a4f64b8 EFLAGS: 00000202 ORIG_RAX: 000000000000002c
[   56.521279] RAX: ffffffffffffffda RBX: 00007ffd7a4f67a8 RCX: 00007f5c619f60d7
[   56.521281] RDX: 0000000000000080 RSI: 00007ffd7a4f64f0 RDI: 0000000000000004
[   56.521282] RBP: 00007ffd7a4f6680 R08: 00007ffd7a4f64d0 R09: 0000000000000014
[   56.521283] R10: 0000000000000000 R11: 0000000000000202 R12: 0000000000000001
[   56.521285] R13: 0000000000000000 R14: 000055c2c648ed58 R15: 00007f5c61b1a000

Quoted 38 lines

Quoted 23 lines
Quoted 10 lines
---
drivers/tty/serial/serial_core.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
drivers/tty/serial/serial_core.c +0 -0
index 2805cad10511..0b2edf185cc7 100644
drivers/tty/serial/serial_core.c static unsigned int uart_write_room(struct tty_struct *tty) · line 647
Quoted 6 lines
drivers/tty/serial/serial_core.c +1 -1
@@ -643,7 +643,10 @@ static unsigned int uart_write_room(struct tty_struct *tty)
 unsigned int ret;
 
 port = uart_port_ref_lock(state, &flags);
- ret = kfifo_avail(&state->port.xmit_fifo);
+ if (!state->port.xmit_buf)
 
This feels odd. What ports have no transmit buffers? And why would this be the only check that is needed for such broken devices?
Maybe let's fix the root cause here, the driver that does not have a transmit buffer at all?
Quoted 6 lines
Do you suggest we should prevent setting line discipline (like N_CAIF) on PORT_UNKNOWN ports? Or should CAIF check the port type before using it? Note that CAIF is currently in orphan status (no active maintainer), so I'm not sure about the process for modifying it. The serial core fix might be more straightforward.
I think you found a real bug here, that is independent of the caif code, and might just be due to the kfifo stuff. See above for my questions here, and if so, your patch is correct, it's just that the Fixes: tag is a bit off.
thanks,
greg k-h