Re: [PATCH] pps: Don't allow PPS_KC_BIND on removed devices
From: Rodolfo Giometti
Date: Mon Jun 01 2026 - 05:43:47 EST
On 30/05/2026 17:00, Calvin Owens wrote:
On Saturday 05/30 at 14:01 +0200, Rodolfo Giometti wrote:
On 30/05/2026 04:46, Calvin Owens wrote:
If userspace holds its file descriptor open, it can call PPS_KC_BIND on
a device which has been unplugged, leaving pps_kc_hardpps_dev as a
dangling pointer after close().
After that sequence, PPS_KC_BIND is broken until the system is rebooted,
because the pointer comparison in pps_kc_bind() can never be true.
calling pps_ktimer_init+0x0/0x1000 [pps_ktimer] @ 1081
initcall pps_ktimer_init+0x0/0x1000 [pps_ktimer] returned 0 after 811 usecs
pps pps0: bound kernel consumer: edge=0x1
pps pps0: unbound kernel consumer on device removal
pps pps0: bound kernel consumer: edge=0x1
calling pps_ktimer_init+0x0/0x1000 [pps_ktimer] @ 1085
initcall pps_ktimer_init+0x0/0x1000 [pps_ktimer] returned 0 after 340 usecs
pps pps0: another kernel consumer is already bound
Here is a short reproducer, which uses rmmod of the pps-ktimer testcase
to simulate a device being unplugged:
#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
#include <fcntl.h>
#include <sys/ioctl.h>
#include <linux/pps.h>
#include <errno.h>
#include <err.h>
int main(void)
{
while (1) {
int fd;
if (system("insmod ./pps-ktimer.ko"))
err(1, "insmod failed");
fd = open("/dev/pps0", O_RDWR);
if (fd == -1)
err(1, "open failed");
struct pps_bind_args args = {
.tsformat = PPS_TSFMT_TSPEC,
.edge = PPS_CAPTUREASSERT,
.consumer = PPS_KC_HARDPPS,
};
if (ioctl(fd, PPS_KC_BIND, &args))
err(1, "first PPS_KC_BIND failed");
if (system("rmmod pps-ktimer"))
err(1, "rmmod failed");
if (ioctl(fd, PPS_KC_BIND, &args)) {
if (errno != ENODEV)
err(1, "second PPS_KC_BIND failed");
else
puts("Got ENODEV, kernel is patched");
}
close(fd);
}
}
Fix this by setting a flag when the device is unplugged, returning
-ENODEV from PPS_KC_BIND if the flag is set.
For userspace to encounter this new behavior, it must do something which
breaks the interface today, so this fix shouldn't cause any observable
behavior change for working programs.
Reported-by: Sashiko <sashiko-bot@xxxxxxxxxx>
Closes: https://sashiko.dev/#/patchset/cover.1779733602.git.calvin%40wbinvd.org?part=1
Signed-off-by: Calvin Owens <calvin@xxxxxxxxxx>
---
I generally hate "just add a flag" as a way to fix this sort of thing,
but I think this really is specific to KC_BIND and not a more general
lifetime problem.
drivers/pps/kc.c | 20 +++++++++++---------
include/linux/pps_kernel.h | 1 +
2 files changed, 12 insertions(+), 9 deletions(-)
diff --git a/drivers/pps/kc.c b/drivers/pps/kc.c
index fbd23295afd7..5f9058f663c8 100644
--- a/drivers/pps/kc.c
+++ b/drivers/pps/kc.c
@@ -36,17 +36,21 @@ static int pps_kc_hardpps_mode; /* mode bits for kernel consumer */
int pps_kc_bind(struct pps_device *pps, struct pps_bind_args *bind_args)
{
/* Check if another consumer is already bound */
- spin_lock_irq(&pps_kc_hardpps_lock);
+ guard(spinlock_irq)(&pps_kc_hardpps_lock);
+
+ /*
+ * Don't allow PPS_KC_BIND on a removed device.
+ */
+ if (pps->kc_removed)
+ return -ENODEV;
if (bind_args->edge == 0)
if (pps_kc_hardpps_dev == pps) {
pps_kc_hardpps_mode = 0;
pps_kc_hardpps_dev = NULL;
- spin_unlock_irq(&pps_kc_hardpps_lock);
dev_info(&pps->dev, "unbound kernel"
" consumer\n");
} else {
- spin_unlock_irq(&pps_kc_hardpps_lock);
dev_err(&pps->dev, "selected kernel consumer"
" is not bound\n");
return -EINVAL;
@@ -56,11 +60,9 @@ int pps_kc_bind(struct pps_device *pps, struct pps_bind_args *bind_args)
pps_kc_hardpps_dev == pps) {
pps_kc_hardpps_mode = bind_args->edge;
pps_kc_hardpps_dev = pps;
- spin_unlock_irq(&pps_kc_hardpps_lock);
dev_info(&pps->dev, "bound kernel consumer: "
"edge=0x%x\n", bind_args->edge);
} else {
- spin_unlock_irq(&pps_kc_hardpps_lock);
dev_err(&pps->dev, "another kernel consumer"
" is already bound\n");
return -EINVAL;
I strongly dislike the idea of putting dev_info() and dev_err() inside a spinlock.
The original code was designed to release the lock before calling
dev_info(). Using guard() here accidentally breaks that good design by
keeping the lock held during the logging calls.
Generally with RT and NBCON upstream, my assumption is that pepole who
care about this have the tools to deal with it. It's all preemptible on
an RT kernel nowadays.
But I don't feel strongly about this, I'm just curious if you or anyone
else on the list have any thoughts about that. If nobody chimes in, I'll
send a v2 with the guard changes dropped in a few days.
The approach of using a kc_removed flag to return -ENODEV is a correct and robust fix for the core problem, improving both stability and userspace API predictability.
I suggest keeping the kc_removed check as an early exit, but reverting to explicit spin_lock_irq/spin_unlock_irq calls to ensure logging happens outside the spinlock protected region, as in the original code. This balances the benefit of your fix with best practices for spinlock usage.
Could you please prepare a v2 patch following this approach?
Ciao,
Rodolfo
--
GNU/Linux Solutions e-mail: giometti@xxxxxxxxxxxx
Linux Device Driver giometti@xxxxxxxx
Embedded Systems phone: +39 349 2432127
UNIX programming