From: Roberto Sassu <[email protected]>
Registering a winch IRQ is racy, an interrupt may occur before the winch is
added to the winch_handlers list.
If that happens, register_winch_irq() adds to that list a winch that is
scheduled to be (or has already been) freed, causing a panic later in
winch_cleanup().
Avoid the race by adding the winch to the winch_handlers list before
registering the IRQ, and rolling back if um_request_irq() fails.
Signed-off-by: Roberto Sassu <[email protected]>
---
arch/um/drivers/line.c | 14 ++++++++------
1 file changed, 8 insertions(+), 6 deletions(-)
diff --git a/arch/um/drivers/line.c b/arch/um/drivers/line.c
index ffc5cb92fa36..d82bc3fdb86e 100644
--- a/arch/um/drivers/line.c
+++ b/arch/um/drivers/line.c
@@ -676,24 +676,26 @@ void register_winch_irq(int fd, int tty_fd, int pid, struct tty_port *port,
goto cleanup;
}
- *winch = ((struct winch) { .list = LIST_HEAD_INIT(winch->list),
- .fd = fd,
+ *winch = ((struct winch) { .fd = fd,
.tty_fd = tty_fd,
.pid = pid,
.port = port,
.stack = stack });
+ spin_lock(&winch_handler_lock);
+ list_add(&winch->list, &winch_handlers);
+ spin_unlock(&winch_handler_lock);
+
if (um_request_irq(WINCH_IRQ, fd, IRQ_READ, winch_interrupt,
IRQF_SHARED, "winch", winch) < 0) {
printk(KERN_ERR "register_winch_irq - failed to register "
"IRQ\n");
+ spin_lock(&winch_handler_lock);
+ list_del(&winch->list);
+ spin_unlock(&winch_handler_lock);
goto out_free;
}
- spin_lock(&winch_handler_lock);
- list_add(&winch->list, &winch_handlers);
- spin_unlock(&winch_handler_lock);
-
return;
out_free:
--
2.34.1
On Thu, 2024-03-07 at 11:49 +0100, Roberto Sassu wrote:
> From: Roberto Sassu <[email protected]>
>
> Registering a winch IRQ is racy, an interrupt may occur before the winch is
> added to the winch_handlers list.
>
> If that happens, register_winch_irq() adds to that list a winch that is
> scheduled to be (or has already been) freed, causing a panic later in
> winch_cleanup().
>
> Avoid the race by adding the winch to the winch_handlers list before
> registering the IRQ, and rolling back if um_request_irq() fails.
>
> Signed-off-by: Roberto Sassu <[email protected]>
Fixes: 42a359e31a0e ("uml: SIGIO support cleanup")
I see that before that commit there was the same ordering (list_add()
before um_request_irq()).
Failure from um_request_irq() should not result in executing
winch_interrupt() which could call list_del() itself. Then, it should
be fine to delete the winch in the error path.
Roberto
> ---
> arch/um/drivers/line.c | 14 ++++++++------
> 1 file changed, 8 insertions(+), 6 deletions(-)
>
> diff --git a/arch/um/drivers/line.c b/arch/um/drivers/line.c
> index ffc5cb92fa36..d82bc3fdb86e 100644
> --- a/arch/um/drivers/line.c
> +++ b/arch/um/drivers/line.c
> @@ -676,24 +676,26 @@ void register_winch_irq(int fd, int tty_fd, int pid, struct tty_port *port,
> goto cleanup;
> }
>
> - *winch = ((struct winch) { .list = LIST_HEAD_INIT(winch->list),
> - .fd = fd,
> + *winch = ((struct winch) { .fd = fd,
> .tty_fd = tty_fd,
> .pid = pid,
> .port = port,
> .stack = stack });
>
> + spin_lock(&winch_handler_lock);
> + list_add(&winch->list, &winch_handlers);
> + spin_unlock(&winch_handler_lock);
> +
> if (um_request_irq(WINCH_IRQ, fd, IRQ_READ, winch_interrupt,
> IRQF_SHARED, "winch", winch) < 0) {
> printk(KERN_ERR "register_winch_irq - failed to register "
> "IRQ\n");
> + spin_lock(&winch_handler_lock);
> + list_del(&winch->list);
> + spin_unlock(&winch_handler_lock);
> goto out_free;
> }
>
> - spin_lock(&winch_handler_lock);
> - list_add(&winch->list, &winch_handlers);
> - spin_unlock(&winch_handler_lock);
> -
> return;
>
> out_free:
On 3/7/2024 1:43 PM, Roberto Sassu wrote:
> On Thu, 2024-03-07 at 11:49 +0100, Roberto Sassu wrote:
>> From: Roberto Sassu <[email protected]>
>>
>> Registering a winch IRQ is racy, an interrupt may occur before the winch is
>> added to the winch_handlers list.
>>
>> If that happens, register_winch_irq() adds to that list a winch that is
>> scheduled to be (or has already been) freed, causing a panic later in
>> winch_cleanup().
>>
>> Avoid the race by adding the winch to the winch_handlers list before
>> registering the IRQ, and rolling back if um_request_irq() fails.
>>
>> Signed-off-by: Roberto Sassu <[email protected]>
>
> Fixes: 42a359e31a0e ("uml: SIGIO support cleanup")
>
> I see that before that commit there was the same ordering (list_add()
> before um_request_irq()).
>
> Failure from um_request_irq() should not result in executing
> winch_interrupt() which could call list_del() itself. Then, it should
> be fine to delete the winch in the error path.
Richard, did you have time to look at this?
Thanks
Roberto
> Roberto
>
>> ---
>> arch/um/drivers/line.c | 14 ++++++++------
>> 1 file changed, 8 insertions(+), 6 deletions(-)
>>
>> diff --git a/arch/um/drivers/line.c b/arch/um/drivers/line.c
>> index ffc5cb92fa36..d82bc3fdb86e 100644
>> --- a/arch/um/drivers/line.c
>> +++ b/arch/um/drivers/line.c
>> @@ -676,24 +676,26 @@ void register_winch_irq(int fd, int tty_fd, int pid, struct tty_port *port,
>> goto cleanup;
>> }
>>
>> - *winch = ((struct winch) { .list = LIST_HEAD_INIT(winch->list),
>> - .fd = fd,
>> + *winch = ((struct winch) { .fd = fd,
>> .tty_fd = tty_fd,
>> .pid = pid,
>> .port = port,
>> .stack = stack });
>>
>> + spin_lock(&winch_handler_lock);
>> + list_add(&winch->list, &winch_handlers);
>> + spin_unlock(&winch_handler_lock);
>> +
>> if (um_request_irq(WINCH_IRQ, fd, IRQ_READ, winch_interrupt,
>> IRQF_SHARED, "winch", winch) < 0) {
>> printk(KERN_ERR "register_winch_irq - failed to register "
>> "IRQ\n");
>> + spin_lock(&winch_handler_lock);
>> + list_del(&winch->list);
>> + spin_unlock(&winch_handler_lock);
>> goto out_free;
>> }
>>
>> - spin_lock(&winch_handler_lock);
>> - list_add(&winch->list, &winch_handlers);
>> - spin_unlock(&winch_handler_lock);
>> -
>> return;
>>
>> out_free:
On Thu, 2024-03-07 at 11:49 +0100, Roberto Sassu wrote:
> From: Roberto Sassu <[email protected]>
>
> Registering a winch IRQ is racy, an interrupt may occur before the winch is
> added to the winch_handlers list.
>
> If that happens, register_winch_irq() adds to that list a winch that is
> scheduled to be (or has already been) freed, causing a panic later in
> winch_cleanup().
>
> Avoid the race by adding the winch to the winch_handlers list before
> registering the IRQ, and rolling back if um_request_irq() fails.
>
Reviewed-by: Johannes Berg <[email protected]>
johannes
On Thu, 2024-03-28 at 09:25 +0100, Johannes Berg wrote:
> On Thu, 2024-03-07 at 11:49 +0100, Roberto Sassu wrote:
> > From: Roberto Sassu <[email protected]>
> >
> > Registering a winch IRQ is racy, an interrupt may occur before the winch is
> > added to the winch_handlers list.
> >
> > If that happens, register_winch_irq() adds to that list a winch that is
> > scheduled to be (or has already been) freed, causing a panic later in
> > winch_cleanup().
> >
> > Avoid the race by adding the winch to the winch_handlers list before
> > registering the IRQ, and rolling back if um_request_irq() fails.
> >
>
> Reviewed-by: Johannes Berg <[email protected]>
Thank you! Richard, are you going to pick this up?
Thanks
Roberto
----- Ursprüngliche Mail -----
> Von: "Roberto Sassu" <[email protected]>
> An: "Johannes Berg" <[email protected]>, "richard" <[email protected]>, "anton ivanov"
> <[email protected]>
> CC: "linux-kernel" <[email protected]>, "linux-um" <[email protected]>, "Roberto Sassu"
> <[email protected]>
> Gesendet: Dienstag, 23. April 2024 09:22:31
> Betreff: Re: [PATCH] um: Add winch to winch_handlers before registering winch IRQ
> On Thu, 2024-03-28 at 09:25 +0100, Johannes Berg wrote:
>> On Thu, 2024-03-07 at 11:49 +0100, Roberto Sassu wrote:
>> > From: Roberto Sassu <[email protected]>
>> >
>> > Registering a winch IRQ is racy, an interrupt may occur before the winch is
>> > added to the winch_handlers list.
>> >
>> > If that happens, register_winch_irq() adds to that list a winch that is
>> > scheduled to be (or has already been) freed, causing a panic later in
>> > winch_cleanup().
>> >
>> > Avoid the race by adding the winch to the winch_handlers list before
>> > registering the IRQ, and rolling back if um_request_irq() fails.
>> >
>>
>> Reviewed-by: Johannes Berg <[email protected]>
>
> Thank you! Richard, are you going to pick this up?
Yes, it's already in my local queue.
Thanks,
//richard