On Sat, Apr 27, 2024 at 12:23 PM Dmitry Baryshkov dmitry.baryshkov@linaro.org wrote:
On Sat, Apr 27, 2024 at 05:50:10PM +0000, Badhri Jagan Sridharan wrote:
typec_register_partner() does not guarantee partner registration to always succeed. In the event of failure, port->partner is set to the error value or NULL. Given that port->partner validity is not checked, this results in the following crash:
[17514.377076][ T275] Unable to handle kernel NULL pointer dereference at virtual address 00000000000003c0 [17514.378270][ T275] pc : run_state_machine+0x1bc8/0x1c08 [17514.378286][ T275] lr : run_state_machine+0x1b90/0x1c08 .. [17514.378377][ T275] Call trace: [17514.378381][ T275] run_state_machine+0x1bc8/0x1c08 [17514.378388][ T275] tcpm_state_machine_work+0x94/0xe4 [17514.378396][ T275] kthread_worker_fn+0x118/0x328 [17514.378406][ T275] kthread+0x1d0/0x23c [17514.378412][ T275] ret_from_fork+0x10/0x20
Nit: thre is little use in the timestamps and the T275 tag. In future patches you can safely ommit such data.
To prevent the crash, check for port->partner validity before derefencing it in all the call sites.
Cc: stable@vger.kernel.org Fixes: c97cd0b4b54e ("usb: typec: tcpm: set initial svdm version based on pd revision") Signed-off-by: Badhri Jagan Sridharan badhri@google.com
drivers/usb/typec/tcpm/tcpm.c | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-)
diff --git a/drivers/usb/typec/tcpm/tcpm.c b/drivers/usb/typec/tcpm/tcpm.c index ab6ed6111ed0..8a4fa8d875c6 100644 --- a/drivers/usb/typec/tcpm/tcpm.c +++ b/drivers/usb/typec/tcpm/tcpm.c @@ -4,7 +4,6 @@
- USB Power Delivery protocol stack.
*/
#include <linux/completion.h> #include <linux/debugfs.h> #include <linux/device.h> @@ -1580,7 +1579,8 @@ static void svdm_consume_identity(struct tcpm_port *port, const u32 *p, int cnt) port->partner_ident.cert_stat = p[VDO_INDEX_CSTAT]; port->partner_ident.product = product;
typec_partner_set_identity(port->partner);
if (!IS_ERR_OR_NULL(port->partner))I think that it might be better to check typec_register_partner() result in tcpm_typec_connect() and skip setting port->partner in case it returned an error (an error message would be also beneficial). See for example, how this is handled in ucsi_register_partner().
Then these checks can be turned into simple `if (port->partner)`.
Sounds good ! However, I was wondering whether the error message is redundant as typec_register_partner() also prints one. I have anyways added that in V3.
Thanks, Badhri
typec_partner_set_identity(port->partner); tcpm_log(port, "Identity: %04x:%04x.%04x", PD_IDH_VID(vdo),@@ -1742,6 +1742,9 @@ static void tcpm_register_partner_altmodes(struct tcpm_port *port) struct typec_altmode *altmode; int i;
if (IS_ERR_OR_NULL(port->partner))return;for (i = 0; i < modep->altmodes; i++) { altmode = typec_partner_register_altmode(port->partner, &modep->altmode_desc[i]);@@ -4244,7 +4247,8 @@ static void tcpm_typec_connect(struct tcpm_port *port) port->partner = typec_register_partner(port->typec_port, &port->partner_desc); port->connected = true;
typec_partner_set_usb_power_delivery(port->partner, port->partner_pd);
if (!IS_ERR_OR_NULL(port->partner))typec_partner_set_usb_power_delivery(port->partner, port->partner_pd); }}
@@ -4323,8 +4327,10 @@ static void tcpm_typec_disconnect(struct tcpm_port *port) port->plug_prime = NULL; port->cable = NULL; if (port->connected) {
typec_partner_set_usb_power_delivery(port->partner, NULL);typec_unregister_partner(port->partner);
if (!IS_ERR_OR_NULL(port->partner)) {typec_partner_set_usb_power_delivery(port->partner, NULL);typec_unregister_partner(port->partner);} port->partner = NULL; port->connected = false; }@@ -4549,6 +4555,9 @@ static enum typec_cc_status tcpm_pwr_opmode_to_rp(enum typec_pwr_opmode opmode)
static void tcpm_set_initial_svdm_version(struct tcpm_port *port) {
if (IS_ERR_OR_NULL(port->partner))return;switch (port->negotiated_rev) { case PD_REV30: break;base-commit: 3f12222a4bebeb13ce06ddecc1610ad32fa835dd
2.44.0.769.g3c40516874-goog
-- With best wishes Dmitry