2022-09-16 07:11:01

by Shengjiu Wang

Subject: [PATCH] rpmsg: char: Avoid double destroy of default endpoint

The rpmsg_dev_remove() in rpmsg_core is the place for releasing
this default endpoint.

So need to avoid destroying the default endpoint in
rpmsg_chrdev_eptdev_destroy(), this should be the same as
rpmsg_eptdev_release(). Otherwise there will be double destroy
issue that ept->refcount report warning:

refcount_t: underflow; use-after-free.

Call trace:
refcount_warn_saturate+0xf8/0x150
virtio_rpmsg_destroy_ept+0xd4/0xec
rpmsg_dev_remove+0x60/0x70

Fixes: bea9b79c2d10 ("rpmsg: char: Add possibility to use default endpoint of the rpmsg device")
Signed-off-by: Shengjiu Wang <[email protected]>
---
drivers/rpmsg/rpmsg_char.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/rpmsg/rpmsg_char.c b/drivers/rpmsg/rpmsg_char.c
index 0850ae34fb88..562d545ac0d3 100644
--- a/drivers/rpmsg/rpmsg_char.c
+++ b/drivers/rpmsg/rpmsg_char.c
@@ -76,7 +76,8 @@ int rpmsg_chrdev_eptdev_destroy(struct device *dev, void *data)

mutex_lock(&eptdev->ept_lock);
if (eptdev->ept) {
- rpmsg_destroy_ept(eptdev->ept);
+ if (!eptdev->default_ept)
+ rpmsg_destroy_ept(eptdev->ept);
eptdev->ept = NULL;
}
mutex_unlock(&eptdev->ept_lock);
--
2.34.1

2022-09-19 21:00:02

by Mathieu Poirier

Subject: Re: [PATCH] rpmsg: char: Avoid double destroy of default endpoint

On Fri, 16 Sept 2022 at 00:57, Shengjiu Wang <[email protected]> wrote:
>
> The rpmsg_dev_remove() in rpmsg_core is the place for releasing
> this default endpoint.
>
> So need to avoid destroying the default endpoint in
> rpmsg_chrdev_eptdev_destroy(), this should be the same as
> rpmsg_eptdev_release(). Otherwise there will be double destroy
> issue that ept->refcount report warning:
>
> refcount_t: underflow; use-after-free.
>
> Call trace:
> refcount_warn_saturate+0xf8/0x150
> virtio_rpmsg_destroy_ept+0xd4/0xec
> rpmsg_dev_remove+0x60/0x70
>
> Fixes: bea9b79c2d10 ("rpmsg: char: Add possibility to use default endpoint of the rpmsg device")
> Signed-off-by: Shengjiu Wang <[email protected]>
> ---
> drivers/rpmsg/rpmsg_char.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/rpmsg/rpmsg_char.c b/drivers/rpmsg/rpmsg_char.c
> index 0850ae34fb88..562d545ac0d3 100644
> --- a/drivers/rpmsg/rpmsg_char.c
> +++ b/drivers/rpmsg/rpmsg_char.c
> @@ -76,7 +76,8 @@ int rpmsg_chrdev_eptdev_destroy(struct device *dev, void *data)
>
> mutex_lock(&eptdev->ept_lock);
> if (eptdev->ept) {
> - rpmsg_destroy_ept(eptdev->ept);
> + if (!eptdev->default_ept)
> + rpmsg_destroy_ept(eptdev->ept);
> eptdev->ept = NULL;

Arnaud - please look into this one.

Thanks,
Mathieu

> }
> mutex_unlock(&eptdev->ept_lock);
> --
> 2.34.1
>

2022-09-20 17:12:14

by Arnaud Pouliquen

Subject: Re: [PATCH] rpmsg: char: Avoid double destroy of default endpoint

Hello Shengjiu,

On 9/16/22 08:39, Shengjiu Wang wrote:
> The rpmsg_dev_remove() in rpmsg_core is the place for releasing
> this default endpoint.
>
> So need to avoid destroying the default endpoint in
> rpmsg_chrdev_eptdev_destroy(), this should be the same as
> rpmsg_eptdev_release(). Otherwise there will be double destroy
> issue that ept->refcount report warning:
>
> refcount_t: underflow; use-after-free.
>
> Call trace:
> refcount_warn_saturate+0xf8/0x150
> virtio_rpmsg_destroy_ept+0xd4/0xec
> rpmsg_dev_remove+0x60/0x70
>
> Fixes: bea9b79c2d10 ("rpmsg: char: Add possibility to use default endpoint of the rpmsg device")
> Signed-off-by: Shengjiu Wang <[email protected]>
> ---
> drivers/rpmsg/rpmsg_char.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/rpmsg/rpmsg_char.c b/drivers/rpmsg/rpmsg_char.c
> index 0850ae34fb88..562d545ac0d3 100644
> --- a/drivers/rpmsg/rpmsg_char.c
> +++ b/drivers/rpmsg/rpmsg_char.c
> @@ -76,7 +76,8 @@ int rpmsg_chrdev_eptdev_destroy(struct device *dev, void *data)
>
> mutex_lock(&eptdev->ept_lock);
> if (eptdev->ept) {
> - rpmsg_destroy_ept(eptdev->ept);
> + if (!eptdev->default_ept)
> + rpmsg_destroy_ept(eptdev->ept);
> eptdev->ept = NULL;
> }
> mutex_unlock(&eptdev->ept_lock);


Good catch!
I tried to reproduce it but without success. Do you set a specific config?

Concerning the patch:

Could you add a comment, to help to understand the condition?
something like "The default endpoint is released by the rpmsg core"

With that Reviewed-by: Arnaud Pouliquen <[email protected]>

Thanks,
Arnaud