Re: [PATCH v2 0/4] PCI: Fix ACS enablement for Root Ports in OF platforms
From: Marek Szyprowski
Date: Fri Dec 12 2025 - 02:25:47 EST
On 12.12.2025 05:02, Manivannan Sadhasivam wrote:
> On Wed, Dec 10, 2025 at 06:26:27PM +0100, Marek Szyprowski wrote:
>> On 09.12.2025 16:04, Manivannan Sadhasivam wrote:
>>> On Tue, Dec 09, 2025 at 01:00:55PM +0100, Marek Szyprowski wrote:
>>>> On 09.12.2025 12:15, Manivannan Sadhasivam wrote:
>>>>> On Tue, Dec 09, 2025 at 09:28:38AM +0100, Marek Szyprowski wrote:
>>>>>> On 09.12.2025 08:31, Marek Szyprowski wrote:
>>>>>>> On 04.12.2025 14:13, Marek Szyprowski wrote:
>>>>>>>> On 03.12.2025 13:04, Marek Szyprowski wrote:
>>>>>>>>> On 02.12.2025 15:22, Manivannan Sadhasivam wrote:
>>>>>>>>>> This series fixes the long standing issue with ACS in OF platforms.
>>>>>>>>>> There are
>>>>>>>>>> two fixes in this series, both fixing independent issues on their
>>>>>>>>>> own, but both
>>>>>>>>>> are needed to properly enable ACS on OF platforms.
>>>>>>>>>>
>>>>>>>>>> Issue(s) background
>>>>>>>>>> ===================
>>>>>>>>>>
>>>>>>>>>> Back in 2021, Xingang Wang first noted a failure in attaching the
>>>>>>>>>> HiSilicon SEC
>>>>>>>>>> device to QEMU ARM64 pci-root-port device [1]. He then tracked down
>>>>>>>>>> the issue to
>>>>>>>>>> ACS not being enabled for the QEMU Root Port device and he proposed
>>>>>>>>>> a patch to
>>>>>>>>>> fix it [2].
>>>>>>>>>>
>>>>>>>>>> Once the patch got applied, people reported PCIe issues with
>>>>>>>>>> linux-next on the
>>>>>>>>>> ARM Juno Development boards, where they saw failure in enumerating
>>>>>>>>>> the endpoint
>>>>>>>>>> devices [3][4]. So soon, the patch got dropped, but the actual
>>>>>>>>>> issue with the
>>>>>>>>>> ARM Juno boards was left behind.
>>>>>>>>>>
>>>>>>>>>> Fast forward to 2024, Pavan resubmitted the same fix [5] for his
>>>>>>>>>> own usecase,
>>>>>>>>>> hoping that someone in the community would fix the issue with ARM
>>>>>>>>>> Juno boards.
>>>>>>>>>> But the patch was rightly rejected, as a patch that was known to
>>>>>>>>>> cause issues
>>>>>>>>>> should not be merged to the kernel. But again, no one investigated
>>>>>>>>>> the Juno
>>>>>>>>>> issue and it was left behind again.
>>>>>>>>>>
>>>>>>>>>> Now it ended up in my plate and I managed to track down the issue
>>>>>>>>>> with the help
>>>>>>>>>> of Naresh who got access to the Juno boards in LKFT. The Juno issue
>>>>>>>>>> was with the
>>>>>>>>>> PCIe switch from Microsemi/IDT, which triggers ACS Source
>>>>>>>>>> Validation error on
>>>>>>>>>> Completions received for the Configuration Read Request from a
>>>>>>>>>> device connected
>>>>>>>>>> to the downstream port that has not yet captured the PCIe bus
>>>>>>>>>> number. As per the
>>>>>>>>>> PCIe spec r6.0 sec 2.2.6.2, "Functions must capture the Bus and
>>>>>>>>>> Device Numbers
>>>>>>>>>> supplied with all Type 0 Configuration Write Requests completed by
>>>>>>>>>> the Function
>>>>>>>>>> and supply these numbers in the Bus and Device Number fields of the
>>>>>>>>>> Requester ID
>>>>>>>>>> for all Requests". So during the first Configuration Read Request
>>>>>>>>>> issued by the
>>>>>>>>>> switch downstream port during enumeration (for reading Vendor ID),
>>>>>>>>>> Bus and
>>>>>>>>>> Device numbers will be unknown to the device. So it responds to the
>>>>>>>>>> Read Request
>>>>>>>>>> with Completion having Bus and Device number as 0. The switch
>>>>>>>>>> interprets the
>>>>>>>>>> Completion as an ACS Source Validation error and drops the
>>>>>>>>>> completion, leading
>>>>>>>>>> to the failure in detecting the endpoint device. Though the PCIe
>>>>>>>>>> spec r6.0, sec
>>>>>>>>>> 6.12.1.1, states that "Completions are never affected by ACS Source
>>>>>>>>>> Validation".
>>>>>>>>>> This behavior is in violation of the spec.
>>>>>>>>>>
>>>>>>>>>> Solution
>>>>>>>>>> ========
>>>>>>>>>>
>>>>>>>>>> In September, I submitted a series [6] to fix both issues. For the
>>>>>>>>>> IDT issue,
>>>>>>>>>> I reused the existing quirk in the PCI core which does a dummy
>>>>>>>>>> config write
>>>>>>>>>> before issuing the first config read to the device. And for the ACS
>>>>>>>>>> enablement
>>>>>>>>>> issue, I just resubmitted the original patch from Xingang which called
>>>>>>>>>> pci_request_acs() from devm_of_pci_bridge_init().
>>>>>>>>>>
>>>>>>>>>> But during the review of the series, several comments were received
>>>>>>>>>> and they
>>>>>>>>>> required the series to be reworked completely. Hence, in this
>>>>>>>>>> version, I've
>>>>>>>>>> incorported the comments as below:
>>>>>>>>>>
>>>>>>>>>> 1. For the ACS enablement issue, I've moved the pci_enable_acs()
>>>>>>>>>> call from
>>>>>>>>>> pci_acs_init() to pci_dma_configure().
>>>>>>>>>>
>>>>>>>>>> 2. For the IDT issue, I've cached the ACS capabilities (RO) in
>>>>>>>>>> 'pci_dev',
>>>>>>>>>> collected the broken capability for the IDT switches in the quirk
>>>>>>>>>> and used it to
>>>>>>>>>> disable the capability in the cache. This also allowed me to get
>>>>>>>>>> rid of the
>>>>>>>>>> earlier workaround for the switch.
>>>>>>>>>>
>>>>>>>>>> [1]
>>>>>>>>>> https://lore.kernel.org/all/038397a6-57e2-b6fc-6e1c-7c03b7be9d96@xxxxxxxxxx
>>>>>>>>>> [2]
>>>>>>>>>> https://lore.kernel.org/all/1621566204-37456-1-git-send-email-wangxingang5@xxxxxxxxxx
>>>>>>>>>> [3]
>>>>>>>>>> https://lore.kernel.org/all/01314d70-41e6-70f9-e496-84091948701a@xxxxxxxxxxx
>>>>>>>>>> [4]
>>>>>>>>>> https://lore.kernel.org/all/CADYN=9JWU3CMLzMEcD5MSQGnaLyDRSKc5SofBFHUax6YuTRaJA@xxxxxxxxxxxxxx
>>>>>>>>>> [5]
>>>>>>>>>> https://lore.kernel.org/linux-pci/20241107-pci_acs_fix-v1-1-185a2462a571@xxxxxxxxxxx
>>>>>>>>>> [6]
>>>>>>>>>> https://lore.kernel.org/linux-pci/20250910-pci-acs-v1-0-fe9adb65ad7d@xxxxxxxxxxxxxxxx
>>>>>>>>>>
>>>>>>>>> Thanks for this patchset! I've tested it on my ARM Juno R1 and it
>>>>>>>>> looks that it almost works fine. This patchset even fixed some
>>>>>>>>> issues with PCI devices probe, as I again see SATA and GBit ethernet
>>>>>>>>> devices, which were missing since Linux v6.14 (it looks that
>>>>>>>>> I've also missed this in my tests).
>>>>>>>>>
>>>>>>>>> # lspci
>>>>>>>>> 00:00.0 PCI bridge: PLDA PCI Express Core Reference Design (rev 01)
>>>>>>>>> 01:00.0 PCI bridge: Integrated Device Technology, Inc. [IDT] Device
>>>>>>>>> 8090 (rev 02)
>>>>>>>>> 02:01.0 PCI bridge: Integrated Device Technology, Inc. [IDT] Device
>>>>>>>>> 8090 (rev 02)
>>>>>>>>> 02:02.0 PCI bridge: Integrated Device Technology, Inc. [IDT] Device
>>>>>>>>> 8090 (rev 02)
>>>>>>>>> 02:03.0 PCI bridge: Integrated Device Technology, Inc. [IDT] Device
>>>>>>>>> 8090 (rev 02)
>>>>>>>>> 02:0c.0 PCI bridge: Integrated Device Technology, Inc. [IDT] Device
>>>>>>>>> 8090 (rev 02)
>>>>>>>>> 02:10.0 PCI bridge: Integrated Device Technology, Inc. [IDT] Device
>>>>>>>>> 8090 (rev 02)
>>>>>>>>> 02:1f.0 PCI bridge: Integrated Device Technology, Inc. [IDT] Device
>>>>>>>>> 8090 (rev 02)
>>>>>>>>> 03:00.0 Mass storage controller: Silicon Image, Inc. SiI 3132 Serial
>>>>>>>>> ATA Raid II Controller (rev 01)
>>>>>>>>> 08:00.0 Ethernet controller: Marvell Technology Group Ltd. 88E8057
>>>>>>>>> PCI-E Gigabit Ethernet Controller
>>>>>>>>>
>>>>>>>>> However there is also a regression. After applying this patchset
>>>>>>>>> system suspend/resume stopped working. This is probably related to
>>>>>>>>> this message:
>>>>>>>>>
>>>>>>>>> pcieport 0000:02:1f.0: Unable to change power state from D0 to
>>>>>>>>> D3hot, device inaccessible
>>>>>>>>>
>>>>>>>>> which appears after calling 'rtcwake -s10 -mmem'. This might not be
>>>>>>>>> related to this patchset, so I probably need to apply it on older
>>>>>>>>> kernel releases and check.
>>>>>>>> Just one more information - I've applied this patchset on top of
>>>>>>>> v6.16 and it works perfectly on ARM Juno R1. SATA and GBit ethernet
>>>>>>>> are visible again and system suspend/resume works too, so the issue
>>>>>>>> with the latter on top of v6.18 seems not to be directly related to
>>>>>>>> $subject patchset. I will try to bisect this issue when I have some
>>>>>>>> spare time.
>>>>>>>>
>>>>>>>> Feel free to add:
>>>>>>>>
>>>>>>>> Tested-by: Marek Szyprowski <m.szyprowski@xxxxxxxxxxx>
>>>>>>> I spent some time analyzing this regression on Juno R1 and found that:
>>>>>>>
>>>>>>> 1. SATA and GBit Ethernet stopped working after commit bcb81ac6ae3c
>>>>>>> ("iommu: Get DT/ACPI parsing into the proper probe path") merged to
>>>>>>> v6.15-rc1.
>>>>>>>
>>>>>>> 2. With $subject patch applied to enable SATA & GBit ethernet again,
>>>>>>> system suspend/resume stopped working after commit f3ac2ff14834
>>>>>>> ("PCI/ASPM: Enable all ClockPM and ASPM states for devicetree
>>>>>>> platforms") merged to v6.18-rc1.
>>>>>>>
>>>>> Yes, this was expected as if you don't disable ACS, it will cause issues in
>>>>> detecting the devices.
>>>>>
>>>>>>> If I got it right, according to the latter commit message, some quirks
>>>>>>> have to be added to fix the suspend/resume issue. Unfortunately I have
>>>>>>> no idea if this is the Juno R1 or the given PCI devices specific issue.
>>>>>> And one more note, commit df5192d9bb0e ("PCI/ASPM: Enable only L0s and
>>>>>> L1 for devicetree platforms") doesn't fix the suspend/resume issue
>>>>>> either (with $subject patchset applied on top of it).
>>>>>>
>>>>> Interesting. Can you do:
>>>>>
>>>>> echo performance > /sys/module/pcie_aspm/parameters/policy
>>>>>
>>>>> and then suspend?
>>>> After the above command, system suspend/resume works again.
>>>>
>>> Ok, so ASPM L0s/L1 seems to be the issue. But I'm not quite sure why it causes
>>> issue during suspend/resume. If the device/controller doesn't play well with
>>> ASPM L0s/L1, it should atleast cause the issue before entering suspend.
>>>
>>> I'm clueless here atm...
>> Definitely something gets broken during suspend, after adding
>> 'no_console_suspend' to kernel command line I see the following messages:
>>
>> # time rtcwake -s10 -mmem
>> rtcwake: wakeup from "mem" using /dev/rtc0 at Wed Dec 10 17:04:12 2025
>> PM: suspend entry (deep)
>> Filesystems sync: 0.001 seconds
>> Freezing user space processes
>> Freezing user space processes completed (elapsed 0.005 seconds)
>> OOM killer disabled.
>> Freezing remaining freezable tasks
>> Freezing remaining freezable tasks completed (elapsed 0.003 seconds)
>> psmouse serio1: Failed to disable mouse on 1c070000.kmi
>> psmouse serio0: Failed to disable mouse on 1c060000.kmi
>> pcieport 0000:02:1f.0: Unable to change power state from D0 to D3hot,
>> device inaccessible
> The device just got blown off the bus at this point. But it is unclear to me why
> it happens though if we enable ASPM L0s/L1. I don't think the firmware has
> gotten the chance to turn off the power to devices.
>
> So maybe some actions that we do in the PCI core during system suspend is
> affecting the device state. But can you try to access the device by doing:
>
> lspci -vvv -s 0000:02:1f.0
>
> before initiating system suspend. Just to make sure if the issue happens during
> suspend or way before that.
Same as before:
root@target:~# lspci -vvv -s 0000:02:1f.0
02:1f.0 PCI bridge: Integrated Device Technology, Inc. [IDT] Device 8090
(rev 02) (prog-if 00 [Normal decode])
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-
ParErr- Stepping- SERR- FastB2B- DisINTx+
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort-
<TAbort- <MAbort- >SERR- <PERR- INTx-
Latency: 0
Interrupt: pin ? routed to IRQ 50
Bus: primary=02, secondary=08, subordinate=08, sec-latency=0
I/O behind bridge: 00002000-00002fff
Memory behind bridge: 50100000-501fffff
Prefetchable memory behind bridge:
00000000fff00000-00000000000fffff
Secondary status: 66MHz- FastB2B- ParErr- DEVSEL=fast >TAbort-
<TAbort- <MAbort- <SERR- <PERR-
BridgeCtl: Parity- SERR+ NoISA- VGA- MAbort- >Reset- FastB2B-
PriDiscTmr- SecDiscTmr- DiscTmrStat- DiscTmrSERREn-
Capabilities: [40] Express (v2) Downstream Port (Slot-), MSI 00
DevCap: MaxPayload 2048 bytes, PhantFunc 0
ExtTag+ RBE+
DevCtl: Report errors: Correctable+ Non-Fatal+ Fatal+
Unsupported+
RlxdOrd- ExtTag+ PhantFunc- AuxPwr- NoSnoop-
MaxPayload 128 bytes, MaxReadReq 128 bytes
DevSta: CorrErr- UncorrErr- FatalErr- UnsuppReq- AuxPwr-
TransPend-
LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L0s L1,
Exit Latency L0s <4us, L1 <4us
ClockPM- Surprise+ LLActRep+ BwNot+ ASPMOptComp-
LnkCtl: ASPM L0s L1 Enabled; Disabled- CommClk+
ExtSynch- ClockPM- AutWidDis- BWInt+ AutBWInt+
LnkSta: Speed 2.5GT/s, Width x1, TrErr- Train- SlotClk+
DLActive+ BWMgmt- ABWMgmt-
DevCap2: Completion Timeout: Not Supported, TimeoutDis-,
LTR-, OBFF Not Supported ARIFwd+
DevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis-,
LTR-, OBFF Disabled ARIFwd-
LnkCtl2: Target Link Speed: 5GT/s, EnterCompliance-
SpeedDis-, Selectable De-emphasis: -6dB
Transmit Margin: Normal Operating Range,
EnterModifiedCompliance- ComplianceSOS-
Compliance De-emphasis: -6dB
LnkSta2: Current De-emphasis Level: -6dB,
EqualizationComplete-, EqualizationPhase1-
EqualizationPhase2-, EqualizationPhase3-,
LinkEqualizationRequest-
Capabilities: [c0] Power Management version 3
Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA
PME(D0+,D1-,D2-,D3hot+,D3cold+)
Status: D0 NoSoftRst+ PME-Enable- DSel=0 DScale=0 PME-
Capabilities: [d0] MSI: Enable+ Count=1/1 Maskable- 64bit+
Address: 00000000fffbb040 Data: 00e7
Capabilities: [100 v2] Advanced Error Reporting
UESta: DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt-
UnxCmplt- RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-
UEMsk: DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt-
UnxCmplt- RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-
UESvrt: DLP+ SDES+ TLP- FCP+ CmpltTO- CmpltAbrt-
UnxCmplt- RxOF+ MalfTLP+ ECRC- UnsupReq- ACSViol-
CESta: RxErr- BadTLP- BadDLLP- Rollover- Timeout-
NonFatalErr-
CEMsk: RxErr- BadTLP- BadDLLP- Rollover- Timeout-
NonFatalErr+
AERCap: First Error Pointer: 00, GenCap+ CGenEn- ChkCap+
ChkEn-
Capabilities: [200 v1] Virtual Channel
Caps: LPEVC=0 RefClk=100ns PATEntryBits=1
Arb: Fixed- WRR32- WRR64- WRR128-
Ctrl: ArbSelect=Fixed
Status: InProgress-
VC0: Caps: PATOffset=00 MaxTimeSlots=1 RejSnoopTrans-
Arb: Fixed+ WRR32- WRR64- WRR128- TWRR128-
WRR256-
Ctrl: Enable+ ID=0 ArbSelect=Fixed TC/VC=ff
Status: NegoPending- InProgress-
Capabilities: [320 v1] Access Control Services
ACSCap: SrcValid+ TransBlk+ ReqRedir+ CmpltRedir+
UpstreamFwd+ EgressCtrl+ DirectTrans+
ACSCtl: SrcValid- TransBlk- ReqRedir+ CmpltRedir+
UpstreamFwd+ EgressCtrl- DirectTrans-
Capabilities: [330 v1] #12
Kernel driver in use: pcieport
root@target:~# time rtcwake -s10 -mmem
rtcwake: wakeup from "mem" using /dev/rtc0 at Fri Dec 12 07:23:50 2025
[ 110.529810] PM: suspend entry (deep)
[ 110.532688] Filesystems sync: 0.001 seconds
[ 110.549590] Freezing user space processes
[ 110.557833] Freezing user space processes completed (elapsed 0.008
seconds)
[ 110.558282] OOM killer disabled.
[ 110.558296] Freezing remaining freezable tasks
[ 110.561602] Freezing remaining freezable tasks completed (elapsed
0.003 seconds)
[ 110.736524] psmouse serio1: Failed to disable mouse on 1c070000.kmi
[ 111.071329] psmouse serio0: Failed to disable mouse on 1c060000.kmi
[ 111.700685] pcieport 0000:02:1f.0: Unable to change power state from
D0 to D3hot, device inaccessible
[ 111.737951] Disabling non-boot CPUs ...
[ 111.757973] psci: CPU5 killed (polled 0 ms)
[ 111.775215] psci: CPU4 killed (polled 0 ms)
[ 111.789725] psci: CPU3 killed (polled 4 ms)
[ 111.800778] psci: CPU2 killed (polled 0 ms)
[ 111.816363] psci: CPU1 killed (polled 0 ms)
(machine never wakes up)
Best regards
--
Marek Szyprowski, PhD
Samsung R&D Institute Poland