linux-stable.git/drivers/pci, branch v6.6.2

Revert "PCI/ASPM: Disable only ASPM_STATE_L1 when driver, disables L1"

2023-11-20T10:59:39+00:00

commit 3cb4f534bac010258b2688395c2f13459a932be9 upstream.

This reverts commit fb097dcd5a28c0a2325632405c76a66777a6bed9.

After fb097dcd5a28 ("PCI/ASPM: Disable only ASPM_STATE_L1 when driver
disables L1"), disabling L1 via pci_disable_link_state(PCIE_LINK_STATE_L1),
then enabling one substate, e.g., L1.1, via sysfs actually enables *all*
the substates.

For example, r8169 disables L1 because of hardware issues on a number of
systems, which implicitly disables the L1.1 and L1.2 substates.

On some systems, L1 and L1.1 work fine, but L1.2 causes missed rx packets.
Enabling L1.1 via the sysfs "aspm_l1_1" attribute unexpectedly enables L1.2
as well as L1.1.

After fb097dcd5a28, pci_disable_link_state(PCIE_LINK_STATE_L1) adds only
ASPM_L1 (but not any of the L1.x substates) to the "aspm_disable" mask:

  --- Before fb097dcd5a28
  +++ After fb097dcd5a28

  # r8169 disables L1:
    pci_disable_link_state(PCIE_LINK_STATE_L1)
  -   disable |= ASPM_L1 | ASPM_L1_1 | ASPM_L1_2 | ...  # disable L1, L1.x
  +   disable |= ASPM_L1                                # disable L1 only

  # write "1" to sysfs "aspm_l1_1" attribute:
    l1_1_aspm
      aspm_attr_store_common(state = ASPM_L1_1)
        disable &= ~ASPM_L1_1              # enable L1.1
        if (state & (ASPM_L1_1 | ...))     # if enabling any substate
          disable &= ~ASPM_L1              # enable L1

  # final state:
  - disable = ASPM_L1_2 | ...              # L1, L1.1 enabled; L1.2 disabled
  + disable = 0                            # L1, L1.1, L1.2 all enabled

Enabling an L1.x substate removes the substate and L1 from the
"aspm_disable" mask.  After fb097dcd5a28, the substates were not added to
the mask when disabling L1, so enabling one substate implicitly enables all
of them.

Revert fb097dcd5a28 so enabling one substate doesn't enable the others.

Link: https://lore.kernel.org/r/c75931ac-7208-4200-9ca1-821629cf5e28@gmail.com
Signed-off-by: Heiner Kallweit 
[bhelgaas: work through example in commit log]
Signed-off-by: Bjorn Helgaas 
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman

PCI: endpoint: Fix double free in __pci_epc_create()

2023-11-20T10:59:23+00:00

[ Upstream commit c9501d268944d6c0475ecb3e740a084a7da9cbfe ]

The pci_epc_release() function frees "epc" so the kfree() on the next line
is a double free.  Drop the redundant free.

Fixes: 7711cbb4862a ("PCI: endpoint: Fix WARN() when an endpoint driver is removed")
Link: https://lore.kernel.org/r/2ce68694-87a7-4c06-b8a4-9870c891b580@moroto.mountain
Signed-off-by: Dan Carpenter 
Signed-off-by: Bjorn Helgaas 
Reviewed-by: Manivannan Sadhasivam 
Signed-off-by: Sasha Levin

PCI: vmd: Correct PCI Header Type Register's multi-function check

2023-11-20T10:59:20+00:00

[ Upstream commit 5827e17d0555b566c32044b0632b46f9f95054fa ]

vmd_domain_reset() attempts to find whether the device may contain multiple
functions by checking 0x80 (Multi-Function Device), however, the hdr_type
variable has already been masked with PCI_HEADER_TYPE_MASK so the check can
never true.

To fix the issue, don't mask the read with PCI_HEADER_TYPE_MASK.

Fixes: 6aab5622296b ("PCI: vmd: Clean up domain before enumeration")
Link: https://lore.kernel.org/r/20231003125300.5541-2-ilpo.jarvinen@linux.intel.com
Signed-off-by: Ilpo Järvinen 
Signed-off-by: Bjorn Helgaas 
Cc: Nirmal Patel 
Signed-off-by: Sasha Levin

PCI: Prevent xHCI driver from claiming AMD VanGogh USB3 DRD device

2023-11-08T10:56:21+00:00

commit 7e6f3b6d2c352b5fde37ce3fed83bdf6172eebd4 upstream.

The AMD VanGogh SoC contains a DesignWare USB3 Dual-Role Device that can be
operated as either a USB Host or a USB Device, similar to on the AMD Nolan
platform.

be6646bfbaec ("PCI: Prevent xHCI driver from claiming AMD Nolan USB3 DRD
device") added a quirk to let the dwc3 driver claim the Nolan device since
it provides more specific support.

Extend that quirk to include the VanGogh SoC USB3 device.

Link: https://lore.kernel.org/r/20230927202212.2388216-1-vi@endrift.com
Signed-off-by: Vicki Pfau 
[bhelgaas: include be6646bfbaec reference, add stable tag]
Signed-off-by: Bjorn Helgaas 
Cc: stable@vger.kernel.org	# v3.19+
Signed-off-by: Greg Kroah-Hartman

Merge tag 'pci-v6.6-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci

2023-10-06T16:11:13+00:00

Pull PCI fixes from Bjorn Helgaas:

 - Fix a qcom register offset that broke IPQ8074 PCIe controller
   enumeration (Sricharan Ramabadhran)

 - Handle interrupt parsing failures when creating a device tree node to
   avoid using uninitialized data (Lizhi Hou)

 - Clean up if adding PCI device node fails when creating a device tree
   node to avoid a memory leak (Lizhi Hou)

 - If a link is down, mark all downstream devices as "disconnected" so
   we don't wait for them on resume (Mika Westerberg)

* tag 'pci-v6.6-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci:
  PCI/PM: Mark devices disconnected if upstream PCIe link is down on resume
  PCI: of: Destroy changeset when adding PCI device node fails
  PCI: of_property: Handle interrupt parsing failures
  PCI: qcom: Fix IPQ8074 enumeration

PCI/PM: Mark devices disconnected if upstream PCIe link is down on resume

2023-09-29T22:42:00+00:00

Mark Blakeney reported that when suspending system with a Thunderbolt
dock connected and then unplugging the dock before resume (which is
pretty normal flow with laptops), resuming takes long time.

What happens is that the PCIe link from the root port to the PCIe switch
inside the Thunderbolt device does not train (as expected, the link is
unplugged):

  pcieport 0000:00:07.2: restoring config space at offset 0x24 (was 0x3bf12001, writing 0x3bf12001)
  pcieport 0000:00:07.0: waiting 100 ms for downstream link
  pcieport 0000:01:00.0: not ready 1023ms after resume; giving up

However, at this point we still try to resume the devices below that
unplugged link:

  pcieport 0000:01:00.0: Unable to change power state from D3cold to D0, device inaccessible
  ...
  pcieport 0000:01:00.0: restoring config space at offset 0x38 (was 0xffffffff, writing 0x0)
  ...
  pcieport 0000:02:02.0: waiting 100 ms for downstream link, after activation

And this is the link from PCIe switch downstream port to the xHCI on the
dock:

  xhci_hcd 0000:03:00.0: not ready 65535ms after resume; giving up
  xhci_hcd 0000:03:00.0: Unable to change power state from D3cold to D0, device inaccessible
  xhci_hcd 0000:03:00.0: restoring config space at offset 0x3c (was 0xffffffff, writing 0x1ff)

This ends up slowing down the resume time considerably. For this reason
mark these devices as disconnected if the link above them did not train
properly.

Fixes: e8b908146d44 ("PCI/PM: Increase wait time after resume")
Link: https://lore.kernel.org/r/20230918053041.1018876-1-mika.westerberg@linux.intel.com
Reported-by: Mark Blakeney 
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217915
Signed-off-by: Mika Westerberg 
Signed-off-by: Bjorn Helgaas 
Reviewed-by: Lukas Wunner 
Cc: stable@vger.kernel.org	# v6.4+

PCI: of: Destroy changeset when adding PCI device node fails

2023-09-29T22:33:51+00:00

Previously of_pci_make_dev_node() leaked a cset if it failed to create a
device node for the PCI device with of_changeset_create_node().

Destroy the cset if of_changeset_create_node() fails.

Fixes: 407d1a51921e ("PCI: Create device tree node for bridge")
Link: https://lore.kernel.org/r/1696007417-42059-1-git-send-email-lizhi.hou@amd.com
Reported-by: Herve Codina 
Closes: https://lore.kernel.org/all/20230911171319.495bb837@bootlin.com/
Signed-off-by: Lizhi Hou 
[bhelgaas: commit log]
Signed-off-by: Bjorn Helgaas 
Reviewed-by: Herve Codina

PCI: of_property: Handle interrupt parsing failures

2023-09-29T22:33:46+00:00

of_pci_prop_intr_map() uses uninitialized addr_sz[] values if
of_irq_parse_raw() fails, which leads to intermittent crashes.

Clear addr_sz[] before use so we never use uninitialized elements.

If no valid IRQs are parsed, don't bother adding the interrupt-map
property.

Fixes: 407d1a51921e ("PCI: Create device tree node for bridge")
Link: https://lore.kernel.org/r/1696007448-42127-1-git-send-email-lizhi.hou@amd.com
Reported-by: Jonathan Cameron 
Closes: https://lore.kernel.org/all/20230911154856.000076c3@Huawei.com/
Reported-by: Herve Codina 
Closes: https://lore.kernel.org/all/20230911171319.495bb837@bootlin.com/
Signed-off-by: Lizhi Hou 
[bhelgaas: commit log, add similar report from Herve]
Signed-off-by: Bjorn Helgaas 
Reviewed-by: Herve Codina

PCI: qcom: Fix IPQ8074 enumeration

2023-09-29T20:47:25+00:00

PARF_SLV_ADDR_SPACE_SIZE_2_3_3 is used by qcom_pcie_post_init_2_3_3().
This PCIe slave address space size register offset is 0x358 but was
incorrectly changed to 0x16c by 39171b33f652 ("PCI: qcom: Remove PCIE20_
prefix from register definitions").

This prevented access to slave address space registers like iATU, etc.,
so the IPQ8074 PCIe controller was not enumerated.

Revert back to the correct 0x358 offset and remove the unused
PARF_SLV_ADDR_SPACE_SIZE_2_3_3.

Fixes: 39171b33f652 ("PCI: qcom: Remove PCIE20_ prefix from register definitions")
Link: https://lore.kernel.org/r/20230919102948.1844909-1-quic_srichara@quicinc.com
Tested-by: Robert Marko 
Signed-off-by: Sricharan Ramabadhran 
[bhelgaas: commit log]
Signed-off-by: Bjorn Helgaas 
Reviewed-by: Manivannan Sadhasivam 
Reviewed-by: Konrad Dybcio 
Cc: stable@vger.kernel.org	# v6.4+

PCI/AER: Export pcie_aer_is_native()

2023-09-11T22:24:16+00:00

Export and move the declaration of pcie_aer_is_native() to a common header
file to be reused by cxl/pci module.

Signed-off-by: Smita Koralahalli 
Acked-by: Bjorn Helgaas 
Reviewed-by: Kuppuswamy Sathyanarayanan 
Reviewed-by: Robert Richter 
Reviewed-by: Jonathan Cameron 
Reviewed-by: Dave Jiang 
Link: https://lore.kernel.org/r/20230823234305.27333-3-Smita.KoralahalliChannabasappa@amd.com
Signed-off-by: Dan Williams