From: Janusz Krzysztofik <janusz.krzysztofik@linux.intel.com>
To: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>,
linux-pci@vger.kernel.org, intel-gfx@lists.freedesktop.org,
Joonas Lahtinen <joonas.lahtinen@linux.intel.com>,
Andi Shyti <andi.shyti@linux.intel.com>,
Krzysztof Karas <krzysztof.karas@intel.com>,
Sebastian Brzezinka <sebastian.brzezinka@intel.com>,
Krzysztof Niemiec <krzysztof.niemiec@intel.com>,
linux-kernel@vger.kernel.org,
Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Subject: Re: [PATCH v2] PCI: Disable not supported reset methods on DG1 0x4905
Date: Tue, 08 Sep 2026 17:27:59 +0200 [thread overview]
Message-ID: <c3f23838e78de7c6300bc4c82088ff738167027b.camel@linux.intel.com> (raw)
In-Reply-To: <ap_nbJWGXWTI8ARq@intel.com>
On Tue, 2026-09-08 at 13:46 +0300, Ville Syrjälä wrote:
> On Mon, Sep 07, 2026 at 10:59:18AM +0200, Janusz Krzysztofik wrote:
> > DRM CI consistently reports failing device reset attempts on Intel DG1
> > discrete graphics adapters. Manual reproduction attempts allowed to
> > narrow down the scope of that problem to a specific revision of that
> > hardware, SKU 8086:4905. It has been confirmed that both reset methods
> > claimed by the device, i.e., flr and bus, suffer from that issue.
>
> IIRC the documented workaround for the issue is some kind of magic
> register write after the FLR to unblock some internal state machine,
> but the CPU isn't allowed to do that write and instead it has to be
> performed by the GuC. That means we'd somehow have to reinitialize
> the GuC very early after the FLR, which sounds rather difficult.
>
Hmm, maybe i915 initialization order could be re-arranged to apply that
workaround via GuC before touching the blocked hardware. If that was
possible then maybe we only needed to use a PCI quirk to block consecutive
FLRs before the workaround is applied in order to prevent a hardware
freeze.
I'm adding Daniele to the loop.
Thanks,
Janusz
> >
> > Since consecutively failing device resets may lead to system hangs,
> > disable both useless reset methods. A respective CI test will then be
> > skipped instead of aborting whole CI jobs.
> >
> > Suggested-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
> > Closes: https://gitlab.freedesktop.org/drm/i915/kernel/-/work_items/11814
> > Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
> > Reviewed-by: Krzysztof Karas <krzysztof.karas@intel.com>
> > Signed-off-by: Janusz Krzysztofik <janusz.krzysztofik@linux.intel.com>
> > ---
> > v2: Fix commit description wording (Andi)
> > - add Suggested-by: tag missing from v1,
> > - add collected Reviewed-by: tags.
> >
> > drivers/pci/quirks.c | 5 +++++
> > 1 file changed, 5 insertions(+)
> >
> > diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> > index de9bbccda21fd..e55475374c6fd 100644
> > --- a/drivers/pci/quirks.c
> > +++ b/drivers/pci/quirks.c
> > @@ -3811,6 +3811,9 @@ DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_TI, 0xb005, quirk_no_bus_reset);
> > */
> > DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ASMEDIA, 0x1164, quirk_no_bus_reset);
> >
> > +/* Intel DG1 0x4905 may not recover from bus reset */
> > +DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x4905, quirk_no_bus_reset);
> > +
> > static void quirk_no_pm_reset(struct pci_dev *dev)
> > {
> > /*
> > @@ -5597,6 +5600,7 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x443, quirk_intel_qat_vf_cap);
> > * AMD Neural Processing Unit 0x1502 0x17f0
> > * Intel 82579LM Gigabit Ethernet Controller 0x1502
> > * Intel 82579V Gigabit Ethernet Controller 0x1503
> > + * Intel DG1 Discrete Graphics Adapter 0x4905
> > * Mediatek MT7922 802.11ax PCI Express Wireless Network Adapter
> > * Mediatek MT7925 802.11be PCI Express Wireless Network Adapter
> > */
> > @@ -5612,6 +5616,7 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_AMD, 0x1502, quirk_no_flr);
> > DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_AMD, 0x17f0, quirk_no_flr);
> > DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x1502, quirk_no_flr);
> > DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x1503, quirk_no_flr);
> > +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x4905, quirk_no_flr);
> > DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_MEDIATEK, 0x0616, quirk_no_flr);
> > DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_MEDIATEK, 0x7925, quirk_no_flr);
> >
> > --
> > 2.54.0
prev parent reply other threads:[~2026-09-08 15:28 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-07 8:59 Janusz Krzysztofik
2026-09-08 10:46 ` Ville Syrjälä
2026-09-08 15:27 ` Janusz Krzysztofik [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=c3f23838e78de7c6300bc4c82088ff738167027b.camel@linux.intel.com \
--to=janusz.krzysztofik@linux.intel.com \
--cc=andi.shyti@linux.intel.com \
--cc=bhelgaas@google.com \
--cc=daniele.ceraolospurio@intel.com \
--cc=intel-gfx@lists.freedesktop.org \
--cc=joonas.lahtinen@linux.intel.com \
--cc=krzysztof.karas@intel.com \
--cc=krzysztof.niemiec@intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pci@vger.kernel.org \
--cc=sebastian.brzezinka@intel.com \
--cc=ville.syrjala@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®