Re: [PATCH 3/4] soc: qcom: pmic_glink: Introduce altmode support

From: Johan Hovold
Date: Tue Oct 25 2022 - 04:33:04 EST


[ Resending to Bjorn's current address. ]

On Wed, Aug 17, 2022 at 08:15:11PM -0700, Bjorn Andersson wrote:
> With the PMIC GLINK service, the host OS subscribes to USB-C altmode
> messages, which are sent by the firmware to notify the host OS about
> state updates and HPD interrupts.
>
> The pmic_glink_altmode driver registers for these notifications and
> propagates the notifications as typec_mux, typec_switch and DRM OOB
> notifications as necessary to implement DisplayPort altmode support.
>
> Signed-off-by: Bjorn Andersson <bjorn.andersson@xxxxxxxxxx>
> ---
> drivers/soc/qcom/Makefile | 1 +
> drivers/soc/qcom/pmic_glink_altmode.c | 477 ++++++++++++++++++++++++++
> 2 files changed, 478 insertions(+)
> create mode 100644 drivers/soc/qcom/pmic_glink_altmode.c

> diff --git a/drivers/soc/qcom/pmic_glink_altmode.c b/drivers/soc/qcom/pmic_glink_altmode.c
> new file mode 100644
> index 000000000000..8d2d563cb756
> --- /dev/null
> +++ b/drivers/soc/qcom/pmic_glink_altmode.c

> +static void pmic_glink_altmode_worker(struct work_struct *work)
> +{
> + struct pmic_glink_altmode_port *alt_port = work_to_altmode_port(work);
> + struct pmic_glink_altmode *altmode = alt_port->altmode;
> +
> + typec_switch_set(alt_port->typec_switch, alt_port->orientation);
> +
> + if (alt_port->svid == USB_TYPEC_DP_SID)
> + pmic_glink_altmode_enable_dp(altmode, alt_port, alt_port->mode,
> + alt_port->hpd_state, alt_port->hpd_irq);
> + else
> + pmic_glink_altmode_enable_usb(altmode, alt_port);
> +
> + if (alt_port->hpd_state)
> + drm_bridge_hpd_notify(&alt_port->bridge, connector_status_connected);
> + else
> + drm_bridge_hpd_notify(&alt_port->bridge, connector_status_disconnected);
> +
> + pmic_glink_altmode_request(altmode, ALTMODE_PAN_ACK, alt_port->index);
> +};

I'm seeing fairly frequent crashes during boot of the X13s due to these
notifications being propagated before things have been fully set up:

[ 16.591910] panel-simple-dp-aux aux-aea0000.displayport-controller: Detected SHP LQ140M1JW48 (0x1511)
[ 16.592142] qcom,fastrpc-cb 1b300000.remoteproc:glink-edge:fastrpc:compute-cb@12: Adding to iommu group 17
[ 16.597644] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000010
[ 16.597653] Mem abort info:
[ 16.597657] ESR = 0x0000000096000004
[ 16.597663] EC = 0x25: DABT (current EL), IL = 32 bits
[ 16.597670] SET = 0, FnV = 0
[ 16.597675] EA = 0, S1PTW = 0
[ 16.597680] FSC = 0x04: level 0 translation fault
[ 16.597686] Data abort info:
[ 16.597689] ISV = 0, ISS = 0x00000004
[ 16.597694] CM = 0, WnR = 0
[ 16.597698] user pgtable: 4k pages, 48-bit VAs, pgdp=0000000106b93000
[ 16.597706] [0000000000000010] pgd=0000000000000000, p4d=0000000000000000
[ 16.597722] Internal error: Oops: 0000000096000004 [#1] PREEMPT SMP
[ 16.597731] Dumping ftrace buffer:
[ 16.597742] (ftrace buffer empty)
[ 16.597744] Modules linked in: fastrpc(+) rpmsg_ctrl qrtr_smd rpmsg_char qcom_battmgr pmic_glink_altmode rtc_pm8xxxr
[ 16.597831] CPU: 0 PID: 389 Comm: kworker/0:3 Not tainted 6.1.0-rc2 #195
[ 16.597838] Hardware name: Qualcomm QRD, BIOS 6.0.220110.BOOT.MXF.1.1-00470-MAKENA-1 01/10/2022
[ 16.597842] Workqueue: events pmic_glink_altmode_worker [pmic_glink_altmode]
[ 16.597864] pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[ 16.597870] pc : drm_kms_helper_hotplug_event+0x1c/0x50
[ 16.597882] lr : drm_kms_helper_hotplug_event+0x18/0x50
[ 16.597887] sp : ffff80000c20bca0
[ 16.597889] x29: ffff80000c20bca0 x28: ffffdba5eadbb000 x27: ffff22a9f6f2dc05
[ 16.597898] x26: ffffdba5eadc0b20 x25: ffffdba5eadd8ca0 x24: 0000000000000000
[ 16.597906] x23: 0000000000000003 x22: ffff22a888526000 x21: 0000000000000002
[ 16.597914] x20: ffff22a88ceed000 x19: ffff22a888526000 x18: 0000000000000020
[ 16.597921] x17: 4d003632323d524f x16: 4a414d00313d4755 x15: 4c50544f48006d72
[ 16.597929] x14: 0000000000000001 x13: 0000000000000040 x12: 0000000000000000
[ 16.597936] x11: 0000000000000000 x10: 0000000000000228 x9 : 0000000000000000
[ 16.597944] x8 : 0000000000000000 x7 : 0000000000000000 x6 : 0000000000062e00
[ 16.597951] x5 : 0000000000000000 x4 : ffff22a9f6f2d290 x3 : 0000000000062f00
[ 16.597959] x2 : 0000000000000000 x1 : 0000000000000000 x0 : 0000000000000000
[ 16.597965] Call trace:
[ 16.597968] drm_kms_helper_hotplug_event+0x1c/0x50
[ 16.597973] drm_bridge_connector_hpd_cb+0xa0/0xc0
[ 16.597983] drm_bridge_hpd_notify+0x40/0x60
[ 16.597990] pmic_glink_altmode_worker+0xc0/0x150 [pmic_glink_altmode]
[ 16.598006] process_one_work+0x288/0x6c0
[ 16.598014] worker_thread+0x74/0x450
[ 16.598019] kthread+0x118/0x120
[ 16.598028] ret_from_fork+0x10/0x20
[ 16.598039] Code: f9000bf3 aa0003f3 97ff22af f9445e60 (f9400801)
[ 16.598043] ---[ end trace 0000000000000000 ]---
[ 16.603424] [drm] Initialized msm 1.9.0 20130625 for ae01000.mdp on minor 0

I've verified that it is the funcs pointer in
drm_kms_helper_hotplug_event() which is NULL and a hack like the below
prevents the crash:

diff --git a/drivers/gpu/drm/drm_probe_helper.c b/drivers/gpu/drm/drm_probe_helper.c
index 69b0b2b9cc1c..d515f5b6f3d5 100644
--- a/drivers/gpu/drm/drm_probe_helper.c
+++ b/drivers/gpu/drm/drm_probe_helper.c
@@ -661,7 +661,9 @@ void drm_kms_helper_hotplug_event(struct drm_device *dev)
{
/* send a uevent + call fbdev */
drm_sysfs_hotplug_event(dev);
- if (dev->mode_config.funcs->output_poll_changed)
+
+ WARN_ON(!dev->mode_config.funcs);
+ if (dev->mode_config.funcs && dev->mode_config.funcs->output_poll_changed)
dev->mode_config.funcs->output_poll_changed(dev);

drm_client_dev_hotplug(dev);

It appears that pointer is set in msm_drm_init(), which suggests events
are being forwarded before the driver is ready.

Johan