linux-kernel - Re: [PATCH v2 3/3] usb: typec: ucsi_glink: drop special handling for CCI

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date: Wed, 10 Apr 2024 09:41:39 +0200
From: "Christian A. Ehrhardt" <lk@...e.de>
To: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
Cc: Heikki Krogerus <heikki.krogerus@...ux.intel.com>,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Neil Armstrong <neil.armstrong@...aro.org>,
	linux-usb@...r.kernel.org, linux-kernel@...r.kernel.org,
	linux-arm-msm@...r.kernel.org
Subject: Re: [PATCH v2 3/3] usb: typec: ucsi_glink: drop special handling for
 CCI_BUSY


Hi Dmitry,

On Wed, Apr 10, 2024 at 01:58:58AM +0300, Dmitry Baryshkov wrote:
> On Tue, 9 Apr 2024 at 22:26, Christian A. Ehrhardt <lk@...e.de> wrote:
> >
> >
> > Hi Dmitry,
> >
> > On Tue, Apr 09, 2024 at 06:29:18PM +0300, Dmitry Baryshkov wrote:
> > > Newer Qualcomm platforms (sm8450+) successfully handle busy state and
> > > send the Command Completion after sending the Busy state. Older devices
> > > have firmware bug and can not continue after sending the CCI_BUSY state,
> > > but the command that leads to CCI_BUSY is already forbidden by the
> > > NO_PARTNER_PDOS quirk.
> > >
> > > Follow other UCSI glue drivers and drop special handling for CCI_BUSY
> > > event. Let the UCSI core properly handle this state.
> > >
> > > Reviewed-by: Heikki Krogerus <heikki.krogerus@...ux.intel.com>
> > > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
> > > ---
> > >  drivers/usb/typec/ucsi/ucsi_glink.c | 10 ++++------
> > >  1 file changed, 4 insertions(+), 6 deletions(-)
> > >
> > > diff --git a/drivers/usb/typec/ucsi/ucsi_glink.c b/drivers/usb/typec/ucsi/ucsi_glink.c
> > > index 9ffea20020e7..fe9b951f5228 100644
> > > --- a/drivers/usb/typec/ucsi/ucsi_glink.c
> > > +++ b/drivers/usb/typec/ucsi/ucsi_glink.c
> > > @@ -176,7 +176,8 @@ static int pmic_glink_ucsi_sync_write(struct ucsi *__ucsi, unsigned int offset,
> > >       left = wait_for_completion_timeout(&ucsi->sync_ack, 5 * HZ);
> > >       if (!left) {
> > >               dev_err(ucsi->dev, "timeout waiting for UCSI sync write response\n");
> > > -             ret = -ETIMEDOUT;
> > > +             /* return 0 here and let core UCSI code handle the CCI_BUSY */
> > > +             ret = 0;
> > >       } else if (ucsi->sync_val) {
> > >               dev_err(ucsi->dev, "sync write returned: %d\n", ucsi->sync_val);
> > >       }
> > > @@ -243,11 +244,8 @@ static void pmic_glink_ucsi_notify(struct work_struct *work)
> > >               ucsi_connector_change(ucsi->ucsi, con_num);
> > >       }
> > >
> > > -     if (ucsi->sync_pending && cci & UCSI_CCI_BUSY) {
> > > -             ucsi->sync_val = -EBUSY;
> > > -             complete(&ucsi->sync_ack);
> > > -     } else if (ucsi->sync_pending &&
> > > -                (cci & (UCSI_CCI_ACK_COMPLETE | UCSI_CCI_COMMAND_COMPLETE))) {
> > > +     if (ucsi->sync_pending &&
> > > +         (cci & (UCSI_CCI_ACK_COMPLETE | UCSI_CCI_COMMAND_COMPLETE))) {
> > >               complete(&ucsi->sync_ack);
> > >       }
> > >  }
> >
> > This handling of the command completion turned out to be racy in
> > the ACPI case: If a normal command was sent one should wait for
> > UCSI_CCI_COMMAND_COMPLETE only. In case of an UCSI_ACK_CC_CI
> > command the completion is indicated by UCSI_CCI_ACK_COMPLETE.
> >
> > While not directly related, a port of this
> >     https://lore.kernel.org/all/20240121204123.275441-3-lk@c--e.de/
> > would nicely fit into this series.
> 
> Ack, I'll take a look.

Thanks.

> However... I can not stop but notice that CCG and STM32 glue drivers
> use the same old approach as we do. Which probably means that they
> might have the same issue.

I did ping the ccg people wrt. this but they have a different
workaround that saves them at least most of the time, so I let
this drop.

> Could you please consider pulling up that
> code into the UCSI driver? Maybe the low-level code really should just
> read/write the messages, leaving all completions and CCI parsing to
> the core layer?

I did consider that but one of the ideas behind the new API for
UCSI backends was that backends can send commands (e.g. as part of
a quirk) even in the middle of a ->sync_write() call. Currently,
I don't really see how to combine this with completion handling
in the UCSI core.

> > I don't have the hardware to do this myself.

I did propose other changes to the API with little respone here:
    https://lore.kernel.org/all/20240218222039.822040-1-lk@c--e.de/
That could possibly be extended to achieve this. But again, that
would require testers for all the backends.


Best regards,
Christian