[REGRESSION] iwlwifi IO_PAGE_FAULT and Master Disable Timed Out

[REGRESSION] iwlwifi IO_PAGE_FAULT and Master Disable Timed Out

[Chris Bainbridge]

Hi,

0eaa1f245ac0 ("wifi: iwlwifi: mvm: don't support the reset handshake for
old firmwares") introduced an intermittent boot regression on my laptop:

[   15.974991] iwlwifi 0000:01:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x000a address=0xbe9f5000 flags=0x0000]
[   15.976184] iwlwifi 0000:01:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x000a address=0xbe9f5400 flags=0x0000]
[   15.978501] iwlwifi 0000:01:00.0: Master Disable Timed Out, 100 usec

The commit only contains:

+       /* Those firmware versions claim to support the fw_reset_handshake
+        * but they are buggy.
+        */
+       if (IWL_UCODE_MAJOR(mvm->fw->ucode_ver) <= 77)
+               trans->conf.fw_reset_handshake = false;

The if condition is evaluating true on my laptop despite using firmware
ty-a0-gf-a0-89.ucode. So I suspect this is not the correct way to check
the firmware version?

Also comment #21 on bugzilla.kernel.org bug #220600 (which this commit
was supposed to fix) reports the same "Master Disable Timed Out" error
(although no IO_PAGE_FAULTS, probably the bad memory access is not
caught without iommu=strict).

#regzbot introduced: 0eaa1f245ac03ed0c6394159360532726f666811

RE: [REGRESSION] iwlwifi IO_PAGE_FAULT and Master Disable Timed Out

[Korenblit, Miriam Rachel]

[-- Attachment #1: Type: text/plain, Size: 1850 bytes --]



> -----Original Message-----
> From: Chris Bainbridge <chris.bainbridge@gmail.com>
> Sent: Friday, July 31, 2026 8:05 PM
> To: Grumbach, Emmanuel <emmanuel.grumbach@intel.com>
> Cc: Berg, Johannes <johannes.berg@intel.com>; Korenblit, Miriam Rachel
> <miriam.rachel.korenblit@intel.com>; linux-wireless@vger.kernel.org;
> regressions@lists.linux.dev
> Subject: [REGRESSION] iwlwifi IO_PAGE_FAULT and Master Disable Timed Out
> 
> Hi,
> 
> 0eaa1f245ac0 ("wifi: iwlwifi: mvm: don't support the reset handshake for old
> firmwares") introduced an intermittent boot regression on my laptop:
> 
> [   15.974991] iwlwifi 0000:01:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT
> domain=0x000a address=0xbe9f5000 flags=0x0000]
> [   15.976184] iwlwifi 0000:01:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT
> domain=0x000a address=0xbe9f5400 flags=0x0000]
> [   15.978501] iwlwifi 0000:01:00.0: Master Disable Timed Out, 100 usec
> 
> The commit only contains:
> 
> +       /* Those firmware versions claim to support the fw_reset_handshake
> +        * but they are buggy.
> +        */
> +       if (IWL_UCODE_MAJOR(mvm->fw->ucode_ver) <= 77)
> +               trans->conf.fw_reset_handshake = false;
> 
> The if condition is evaluating true on my laptop despite using firmware ty-a0-gf-
> a0-89.ucode. So I suspect this is not the correct way to check the firmware
> version?

Yes, seems like you are right.
Thanks for the report!

I am attaching a fix candidate, I'd appreciate if you could test it.
> 
> Also comment #21 on bugzilla.kernel.org bug #220600 (which this commit was
> supposed to fix) reports the same "Master Disable Timed Out" error (although no
> IO_PAGE_FAULTS, probably the bad memory access is not caught without
> iommu=strict).
> 
> #regzbot introduced: 0eaa1f245ac03ed0c6394159360532726f666811

[-- Attachment #2: 0001-wifi-iwlwifi-mvm-properly-check-the-FW-version.patch --]
[-- Type: application/octet-stream, Size: 1807 bytes --]

From 825098f2de2bc63e5ce180213d9fad040f644fb2 Mon Sep 17 00:00:00 2001
From: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Date: Sun, 2 Aug 2026 10:16:13 +0300
Subject: [PATCH] wifi: iwlwifi: mvm: properly check the FW version
Organization: Intel Israel (74) Limited

The mentioned commit aimed to disable fw reset handshake for FWs which
their API version is less than 77.
But the check is incorrect: in the old format, we should use
IWL_UCODE_API, and for the new TLV format - fw.ucode_ver.
Since iwlmvm handles only FWs with the new format, we can always use
here fw.ucode_ver.

Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
---
 drivers/net/wireless/intel/iwlwifi/mvm/ops.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/ops.c b/drivers/net/wireless/intel/iwlwifi/mvm/ops.c
index 2297392db955..7e45808da5b3 100644
--- a/drivers/net/wireless/intel/iwlwifi/mvm/ops.c
+++ b/drivers/net/wireless/intel/iwlwifi/mvm/ops.c
@@ -1412,15 +1412,13 @@ iwl_op_mode_mvm_start(struct iwl_trans *trans, const struct iwl_rf_cfg *cfg,
 		 sizeof(mvm->hw->wiphy->fw_version),
 		 "%.31s", fw->fw_version);
 
-	trans->conf.fw_reset_handshake =
-		fw_has_capa(&mvm->fw->ucode_capa,
-			    IWL_UCODE_TLV_CAPA_FW_RESET_HANDSHAKE);
-
-	/* Those firmware versions claim to support the fw_reset_handshake
+	/* Some firmware versions claim to support the fw_reset_handshake
 	 * but they are buggy.
 	 */
-	if (IWL_UCODE_MAJOR(mvm->fw->ucode_ver) <= 77)
-		trans->conf.fw_reset_handshake = false;
+	trans->conf.fw_reset_handshake =
+		fw_has_capa(&mvm->fw->ucode_capa,
+			    IWL_UCODE_TLV_CAPA_FW_RESET_HANDSHAKE) &&
+		mvm->fw->ucode_ver > 77;
 
 	trans->conf.queue_alloc_cmd_ver =
 		iwl_fw_lookup_cmd_ver(mvm->fw,
-- 
2.34.1

Re: [REGRESSION] iwlwifi IO_PAGE_FAULT and Master Disable Timed Out

[Chris Bainbridge]

On Sun, Aug 02, 2026 at 07:24:04AM +0000, Korenblit, Miriam Rachel wrote:
> 
> 
> > -----Original Message-----
> > From: Chris Bainbridge <chris.bainbridge@gmail.com>
> > Sent: Friday, July 31, 2026 8:05 PM
> > To: Grumbach, Emmanuel <emmanuel.grumbach@intel.com>
> > Cc: Berg, Johannes <johannes.berg@intel.com>; Korenblit, Miriam Rachel
> > <miriam.rachel.korenblit@intel.com>; linux-wireless@vger.kernel.org;
> > regressions@lists.linux.dev
> > Subject: [REGRESSION] iwlwifi IO_PAGE_FAULT and Master Disable Timed Out
> > 
> > Hi,
> > 
> > 0eaa1f245ac0 ("wifi: iwlwifi: mvm: don't support the reset handshake for old
> > firmwares") introduced an intermittent boot regression on my laptop:
> > 
> > [   15.974991] iwlwifi 0000:01:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT
> > domain=0x000a address=0xbe9f5000 flags=0x0000]
> > [   15.976184] iwlwifi 0000:01:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT
> > domain=0x000a address=0xbe9f5400 flags=0x0000]
> > [   15.978501] iwlwifi 0000:01:00.0: Master Disable Timed Out, 100 usec
> > 
> > The commit only contains:
> > 
> > +       /* Those firmware versions claim to support the fw_reset_handshake
> > +        * but they are buggy.
> > +        */
> > +       if (IWL_UCODE_MAJOR(mvm->fw->ucode_ver) <= 77)
> > +               trans->conf.fw_reset_handshake = false;
> > 
> > The if condition is evaluating true on my laptop despite using firmware ty-a0-gf-
> > a0-89.ucode. So I suspect this is not the correct way to check the firmware
> > version?
> 
> Yes, seems like you are right.
> Thanks for the report!
> 
> I am attaching a fix candidate, I'd appreciate if you could test it.

Thanks, the fix is working fine.

Reported-and-tested-by: Chris Bainbridge <chris.bainbridge@gmail.com>

Re: [REGRESSION] iwlwifi IO_PAGE_FAULT and Master Disable Timed Out

[Thorsten Leemhuis]

On 8/2/26 09:24, Korenblit, Miriam Rachel wrote:
>> -----Original Message-----
>> From: Chris Bainbridge <chris.bainbridge@gmail.com>
>> Sent: Friday, July 31, 2026 8:05 PM
>> To: Grumbach, Emmanuel <emmanuel.grumbach@intel.com>
>> Cc: Berg, Johannes <johannes.berg@intel.com>; Korenblit, Miriam Rachel
>> <miriam.rachel.korenblit@intel.com>; linux-wireless@vger.kernel.org;
>> regressions@lists.linux.dev
>> Subject: [REGRESSION] iwlwifi IO_PAGE_FAULT and Master Disable Timed Out
>>
>> 0eaa1f245ac0 ("wifi: iwlwifi: mvm: don't support the reset handshake for old
>> firmwares") introduced an intermittent boot regression on my laptop:
>> [...]
> 
> Yes, seems like you are right.
> Thanks for the report!
> 
> I am attaching a fix candidate, I'd appreciate if you could test it.
Miri, Chris, what happened to this regression fix? From the outside it
looks like it fell through the cracks, but well, it's easy to miss
something from there, that's why I'm asking.

Ciao, Thorsten

> From 825098f2de2bc63e5ce180213d9fad040f644fb2 Mon Sep 17 00:00:00 2001
> From: Miri Korenblit <miriam.rachel.korenblit@intel.com>
> Date: Sun, 2 Aug 2026 10:16:13 +0300
> Subject: [PATCH] wifi: iwlwifi: mvm: properly check the FW version
> Organization: Intel Israel (74) Limited
> 
> The mentioned commit aimed to disable fw reset handshake for FWs which
> their API version is less than 77.
> But the check is incorrect: in the old format, we should use
> IWL_UCODE_API, and for the new TLV format - fw.ucode_ver.
> Since iwlmvm handles only FWs with the new format, we can always use
> here fw.ucode_ver.
> 
> Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
> [...]