When ipmi_register_smi fails, it performs a small cleanup routine
before returning its error value. In try_smi_init, on the condition that
ipmi_register_smi fails, ipmi_unregister_smi is called. ipmi_unregister_smi
performs the same cleanup routine as ipmi_register_smi. This results in
a kernel NULL pointer dereference. Removing the cleanup routine in
ipmi_register_smi results in proper cleanup of a ipmi_register_smi failure.
Cc: Corey Minyard <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Jeremy Kerr <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: Russ Anderson <[email protected]>
Acked-by: Andrew Banman <[email protected]>
Signed-off-by: Justin Ernst <[email protected]>
---
drivers/char/ipmi/ipmi_msghandler.c | 10 +++++-----
1 file changed, 5 insertions(+), 5 deletions(-)
diff --git a/drivers/char/ipmi/ipmi_msghandler.c b/drivers/char/ipmi/ipmi_msghandler.c
index 51832b8a2c62..3b0b50c4f064 100644
--- a/drivers/char/ipmi/ipmi_msghandler.c
+++ b/drivers/char/ipmi/ipmi_msghandler.c
@@ -3395,12 +3395,12 @@ int ipmi_register_smi(const struct ipmi_smi_handlers *handlers,
out:
if (rv) {
- ipmi_bmc_unregister(intf);
- list_del_rcu(&intf->link);
+ /*
+ * ipmi_unregister_smi must be called to clean up after
+ * failure. We unlock the mutex to allow ipmi_unregister_smi
+ * to lock it and perform cleanup.
+ */
mutex_unlock(&ipmi_interfaces_mutex);
- synchronize_srcu(&ipmi_interfaces_srcu);
- cleanup_srcu_struct(&intf->users_srcu);
- kref_put(&intf->refcount, intf_free);
} else {
/*
* Keep memory order straight for RCU readers. Make
--
2.12.3
On 08/21/2018 10:25 AM, Justin Ernst wrote:
> When ipmi_register_smi fails, it performs a small cleanup routine
> before returning its error value. In try_smi_init, on the condition that
> ipmi_register_smi fails, ipmi_unregister_smi is called. ipmi_unregister_smi
> performs the same cleanup routine as ipmi_register_smi. This results in
> a kernel NULL pointer dereference. Removing the cleanup routine in
> ipmi_register_smi results in proper cleanup of a ipmi_register_smi failure.
This is almost certainly wrong. If ipmi_register_smi() fails, the
calling code shouldn't call
ipmi_unregister_smi().
However, I think I know what is going wrong. ipmi_register_smi() can
fail after the interface
is initialized, if so the registering code will think it is registered
and call ipmi_unregister_smi()
on a failure. I'll send a patch for this.
-corey
>
> Cc: Corey Minyard <[email protected]>
> Cc: Arnd Bergmann <[email protected]>
> Cc: Greg Kroah-Hartman <[email protected]>
> Cc: Jeremy Kerr <[email protected]>
> Cc: [email protected]
> Cc: [email protected]
> Cc: Russ Anderson <[email protected]>
> Acked-by: Andrew Banman <[email protected]>
> Signed-off-by: Justin Ernst <[email protected]>
> ---
> drivers/char/ipmi/ipmi_msghandler.c | 10 +++++-----
> 1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/char/ipmi/ipmi_msghandler.c b/drivers/char/ipmi/ipmi_msghandler.c
> index 51832b8a2c62..3b0b50c4f064 100644
> --- a/drivers/char/ipmi/ipmi_msghandler.c
> +++ b/drivers/char/ipmi/ipmi_msghandler.c
> @@ -3395,12 +3395,12 @@ int ipmi_register_smi(const struct ipmi_smi_handlers *handlers,
>
> out:
> if (rv) {
> - ipmi_bmc_unregister(intf);
> - list_del_rcu(&intf->link);
> + /*
> + * ipmi_unregister_smi must be called to clean up after
> + * failure. We unlock the mutex to allow ipmi_unregister_smi
> + * to lock it and perform cleanup.
> + */
> mutex_unlock(&ipmi_interfaces_mutex);
> - synchronize_srcu(&ipmi_interfaces_srcu);
> - cleanup_srcu_struct(&intf->users_srcu);
> - kref_put(&intf->refcount, intf_free);
> } else {
> /*
> * Keep memory order straight for RCU readers. Make
> -----Original Message-----
> From: Corey Minyard [mailto:[email protected]]
> Sent: Wednesday, August 22, 2018 11:28 AM
> To: Ernst, Justin <[email protected]>; Corey Minyard
> <[email protected]>
> Cc: Arnd Bergmann <[email protected]>; Greg Kroah-Hartman
> <[email protected]>; Jeremy Kerr <[email protected]>; openipmi-
> [email protected]; [email protected]; Anderson,
> Russ <[email protected]>
> Subject: Re: [PATCH] Remove redundant cleanup in ipmi_register_smi
>
> On 08/21/2018 10:25 AM, Justin Ernst wrote:
> > When ipmi_register_smi fails, it performs a small cleanup routine
> > before returning its error value. In try_smi_init, on the condition
> > that ipmi_register_smi fails, ipmi_unregister_smi is called.
> > ipmi_unregister_smi performs the same cleanup routine as
> > ipmi_register_smi. This results in a kernel NULL pointer dereference.
> > Removing the cleanup routine in ipmi_register_smi results in proper
> cleanup of a ipmi_register_smi failure.
>
> This is almost certainly wrong. If ipmi_register_smi() fails, the calling code
> shouldn't call ipmi_unregister_smi().
>
> However, I think I know what is going wrong. ipmi_register_smi() can fail
> after the interface is initialized, if so the registering code will think it is
> registered and call ipmi_unregister_smi() on a failure. I'll send a patch for
> this.
>
> -corey
Thanks for the reply Corey. I agree with what you have stated. I'll be watching for the patch.
-Justin
>
> >
> > Cc: Corey Minyard <[email protected]>
> > Cc: Arnd Bergmann <[email protected]>
> > Cc: Greg Kroah-Hartman <[email protected]>
> > Cc: Jeremy Kerr <[email protected]>
> > Cc: [email protected]
> > Cc: [email protected]
> > Cc: Russ Anderson <[email protected]>
> > Acked-by: Andrew Banman <[email protected]>
> > Signed-off-by: Justin Ernst <[email protected]>
> > ---
> > drivers/char/ipmi/ipmi_msghandler.c | 10 +++++-----
> > 1 file changed, 5 insertions(+), 5 deletions(-)
> >
> > diff --git a/drivers/char/ipmi/ipmi_msghandler.c
> > b/drivers/char/ipmi/ipmi_msghandler.c
> > index 51832b8a2c62..3b0b50c4f064 100644
> > --- a/drivers/char/ipmi/ipmi_msghandler.c
> > +++ b/drivers/char/ipmi/ipmi_msghandler.c
> > @@ -3395,12 +3395,12 @@ int ipmi_register_smi(const struct
> > ipmi_smi_handlers *handlers,
> >
> > out:
> > if (rv) {
> > - ipmi_bmc_unregister(intf);
> > - list_del_rcu(&intf->link);
> > + /*
> > + * ipmi_unregister_smi must be called to clean up after
> > + * failure. We unlock the mutex to allow ipmi_unregister_smi
> > + * to lock it and perform cleanup.
> > + */
> > mutex_unlock(&ipmi_interfaces_mutex);
> > - synchronize_srcu(&ipmi_interfaces_srcu);
> > - cleanup_srcu_struct(&intf->users_srcu);
> > - kref_put(&intf->refcount, intf_free);
> > } else {
> > /*
> > * Keep memory order straight for RCU readers. Make
>