[Intel-gfx] [char-misc-next 3/4] mei: pxp: re-enable client on errors
Ville Syrjälä
ville.syrjala at linux.intel.com
Tue Nov 14 14:00:27 UTC 2023
On Wed, Oct 11, 2023 at 02:01:56PM +0300, Tomas Winkler wrote:
> From: Alexander Usyskin <alexander.usyskin at intel.com>
>
> Disable and enable mei-pxp client on errors to clean the internal state.
This broke i915 on my Alderlake-P laptop.
Trying to start Xorg just hangs and I eventually have to power off the
laptop to get things back into shape.
The behaviour gets a bit better after commit fb99e79ee62a ("mei: update mei-pxp's
component interface with timeouts") as Xorg "only" gets blocked for
~10 seconds, after which it manages to start, and I get a bunch of spew
in dmesg:
[ 25.431535] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 30.435241] mei_pxp 0000:00:16.0-fbf6fcf1-96cf-4e2e-a6a6-1bab8cbe36b1: Trying to reset the channel...
[ 30.435965] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 30.437341] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 30.437356] i915 0000:00:02.0: [drm] *ERROR* Failed to send tee msg for inv-stream-key-15, ret=[28]
[ 35.555210] mei_pxp 0000:00:16.0-fbf6fcf1-96cf-4e2e-a6a6-1bab8cbe36b1: Trying to reset the channel...
[ 35.555919] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 35.555937] i915 0000:00:02.0: [drm] *ERROR* Failed to send tee msg init arb session, ret=[-62]
[ 35.555941] i915 0000:00:02.0: [drm] *ERROR* tee cmd for arb session creation failed
[ 35.556765] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 36.021808] fuse: init (API version 7.39)
[ 40.675183] mei_pxp 0000:00:16.0-fbf6fcf1-96cf-4e2e-a6a6-1bab8cbe36b1: Trying to reset the channel...
[ 40.676045] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 40.676591] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 40.676602] i915 0000:00:02.0: [drm] *ERROR* Failed to send tee msg for inv-stream-key-15, ret=[28]
[ 40.960209] mate-session-ch[5936]: memfd_create() called without MFD_EXEC or MFD_NOEXEC_SEAL set
[ 45.795172] mei_pxp 0000:00:16.0-fbf6fcf1-96cf-4e2e-a6a6-1bab8cbe36b1: Trying to reset the channel...
[ 45.795872] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 45.796520] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 50.915183] mei_pxp 0000:00:16.0-fbf6fcf1-96cf-4e2e-a6a6-1bab8cbe36b1: Trying to reset the channel...
[ 50.916005] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 50.916012] i915 0000:00:02.0: [drm] *ERROR* Failed to send tee msg for inv-stream-key-15, ret=[-62]
[ 50.916846] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 56.035149] mei_pxp 0000:00:16.0-fbf6fcf1-96cf-4e2e-a6a6-1bab8cbe36b1: Trying to reset the channel...
[ 56.035956] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 56.036585] i915 0000:00:02.0: [drm] *ERROR* Failed to send PXP TEE message
[ 56.036592] i915 0000:00:02.0: [drm] *ERROR* Failed to send tee msg for inv-stream-key-15, ret=[28]
[ 61.155137] mei_pxp 0000:00:16.0-fbf6fcf1-96cf-4e2e-a6a6-1bab8cbe36b1: Trying to reset the channel...
The same spew repeats every time I run any application that uses the GPU,
and the application also gets blocked for a long time (eg. firefox takes
over 15 seconds to start now).
>
> Signed-off-by: Alexander Usyskin <alexander.usyskin at intel.com>
> Signed-off-by: Tomas Winkler <tomas.winkler at intel.com>
> ---
> drivers/misc/mei/pxp/mei_pxp.c | 70 +++++++++++++++++++++++-----------
> 1 file changed, 48 insertions(+), 22 deletions(-)
>
> diff --git a/drivers/misc/mei/pxp/mei_pxp.c b/drivers/misc/mei/pxp/mei_pxp.c
> index c6cdd6a47308ebcc72f34c38..9875d16445bb03efcfb31cd9 100644
> --- a/drivers/misc/mei/pxp/mei_pxp.c
> +++ b/drivers/misc/mei/pxp/mei_pxp.c
> @@ -23,6 +23,24 @@
>
> #include "mei_pxp.h"
>
> +static inline int mei_pxp_reenable(const struct device *dev, struct mei_cl_device *cldev)
> +{
> + int ret;
> +
> + dev_warn(dev, "Trying to reset the channel...\n");
> + ret = mei_cldev_disable(cldev);
> + if (ret < 0)
> + dev_warn(dev, "mei_cldev_disable failed. %d\n", ret);
> + /*
> + * Explicitly ignoring disable failure,
> + * enable may fix the states and succeed
> + */
> + ret = mei_cldev_enable(cldev);
> + if (ret < 0)
> + dev_err(dev, "mei_cldev_enable failed. %d\n", ret);
> + return ret;
> +}
> +
> /**
> * mei_pxp_send_message() - Sends a PXP message to ME FW.
> * @dev: device corresponding to the mei_cl_device
> @@ -35,6 +53,7 @@ mei_pxp_send_message(struct device *dev, const void *message, size_t size)
> {
> struct mei_cl_device *cldev;
> ssize_t byte;
> + int ret;
>
> if (!dev || !message)
> return -EINVAL;
> @@ -44,10 +63,20 @@ mei_pxp_send_message(struct device *dev, const void *message, size_t size)
> byte = mei_cldev_send(cldev, message, size);
> if (byte < 0) {
> dev_dbg(dev, "mei_cldev_send failed. %zd\n", byte);
> - return byte;
> + switch (byte) {
> + case -ENOMEM:
> + fallthrough;
> + case -ENODEV:
> + fallthrough;
> + case -ETIME:
> + ret = mei_pxp_reenable(dev, cldev);
> + if (ret)
> + byte = ret;
> + break;
> + }
> }
>
> - return 0;
> + return byte;
> }
>
> /**
> @@ -63,6 +92,7 @@ mei_pxp_receive_message(struct device *dev, void *buffer, size_t size)
> struct mei_cl_device *cldev;
> ssize_t byte;
> bool retry = false;
> + int ret;
>
> if (!dev || !buffer)
> return -EINVAL;
> @@ -73,26 +103,22 @@ mei_pxp_receive_message(struct device *dev, void *buffer, size_t size)
> byte = mei_cldev_recv(cldev, buffer, size);
> if (byte < 0) {
> dev_dbg(dev, "mei_cldev_recv failed. %zd\n", byte);
> - if (byte != -ENOMEM)
> - return byte;
> -
> - /* Retry the read when pages are reclaimed */
> - msleep(20);
> - if (!retry) {
> - retry = true;
> - goto retry;
> - } else {
> - dev_warn(dev, "No memory on data receive after retry, trying to reset the channel...\n");
> - byte = mei_cldev_disable(cldev);
> - if (byte < 0)
> - dev_warn(dev, "mei_cldev_disable failed. %zd\n", byte);
> - /*
> - * Explicitly ignoring disable failure,
> - * enable may fix the states and succeed
> - */
> - byte = mei_cldev_enable(cldev);
> - if (byte < 0)
> - dev_err(dev, "mei_cldev_enable failed. %zd\n", byte);
> + switch (byte) {
> + case -ENOMEM:
> + /* Retry the read when pages are reclaimed */
> + msleep(20);
> + if (!retry) {
> + retry = true;
> + goto retry;
> + }
> + fallthrough;
> + case -ENODEV:
> + fallthrough;
> + case -ETIME:
> + ret = mei_pxp_reenable(dev, cldev);
> + if (ret)
> + byte = ret;
> + break;
> }
> }
>
> --
> 2.41.0
>
--
Ville Syrjälä
Intel
More information about the Intel-gfx
mailing list