cancel
Showing results for 
Search instead for 
Did you mean: 
Reply
Highlighted
marvinda
Paper Tape
Posts: 5
Registered: ‎09-11-2018
Location: DE
Views: 775
Message 141 of 200

Re: ThinkPad E485/E585 - Firmware bug ACPI IVRS table

This issue (which includes serious IOMMU issues) was now reported more than half a year ago. We got 140 replies, but not a single one from Lenovo. IMO, this situation is telling concerning the state of the ThinkPad brand.

 

At least the A485 received BIOS updates this year. Would anyone know whether the issues are fixed there?

MariuszM
Paper Tape
Posts: 2
Registered: ‎10-27-2018
Location: PL
Views: 768
Message 142 of 200

Re: ThinkPad E485/E585 - Firmware bug ACPI IVRS table

@marvinda I will vote using my wallet. ThinkPad E485 was the last laptop from lenovo.

750lux
Paper Tape
Posts: 9
Registered: ‎03-28-2016
Location: US
Views: 706
Message 143 of 200

Re: ThinkPad E485/E585 - Firmware bug ACPI IVRS table

I bought a e585, this was my first Thinkpad. I really wanted to like it but Lenovo is really dropping the ball on this issue.

laurencedeclan
Fanfold Paper
Posts: 5
Registered: ‎03-01-2019
Location: AU
Views: 662
Message 144 of 200

Re: ThinkPad E485/E585 - Firmware bug ACPI IVRS table

This appears to be an off-by-one error; as in, the relevant IVRS table entries apparently use the indices #33 and #34 instead of the correct #32 and #33.

 

The parameter

ivrs_ioapic[x]=y

allows us to reassign the values correctly, in theory.

 

However:

  • it is not currently possible to blank an entry using the kernel command-line override, so the incorrect entry for #34 likely remains;
  • the override for #33 seems to be ignored by the kernel because the entry for it already exists, which is possibly the cause of the PCIe ASPM issues and the inability to resume from suspend.

Please correct me if I am wrong.

 

From what I can gather on various forums, the A485 had this issue upon release, and it has indeed since been fixed, although I cannot find any references to this in the changelog. Hanlon's razor aside, it's plausible that this remains unfixed so that business clients choose the (more expensive) A485, but I think that Lenovo are simply doing the minimum necessary in terms of bugfixing, and this is not a "Linux-certified" laptop. Additionally, this is a community forum, not a venue for official support, so the amount of traffic here really is irrelevant. How many people in this thread have actually been on the phone with Lenovo over this issue? Perhaps it should be mentioned that this issue affects Windows users as well (presumably this is a "Windows-certified" laptop), at least for those that need a working IOMMU.

 

Let's look at possible solutions. Does anyone know whether this ThinkPad's BIOS images require a digital signature to be flashed? If not, it should be trivial for us to patch this ourselves and utilise the corrected image, sans Lenovo. Alternatively, if the shortcomings in the kernel IVRS override system mentioned above are fixed (open an issue in the kernel bugtracker?), it should be indistinguishable from a patched BIOS. This is all assuming it really is an off-by-one error, of course.

MarcosRC
Fanfold Paper
Posts: 7
Registered: ‎12-18-2018
Location: AU
Views: 644
Message 145 of 200

Re: ThinkPad E485/E585 - Firmware bug ACPI IVRS table

I actually did open a ticket with Lenovo and pointed to this thread. They told me they can't fix it because AMD hasn't sent them the correct details.

 

Others here had opened a ticket with AMD, and they told them it is on Lenovo to set it correctly in the BIOS...

 

In my ticket with Lenovo I pointed out that surely Lenovo has enough clout with AMD to just ask if something is missing instead of sending customers from one company to the other... They just closed the ticket.

 

I had expected much better from Lenovo when I switched from buying ASUS, but never again if this is not solved ASAP. As mentioned, it is a trivial fix, perhaps teen minutes of work for their engineers...

 

tuxlinuxien
Paper Tape
Posts: 2
Registered: ‎09-26-2018
Location: FR
Views: 640
Message 146 of 200

Re: ThinkPad E485/E585 - Firmware bug ACPI IVRS table

I wont say that I am not disappointed by Lenovo but "perhaps teen minutes of work for their engineers" I wonder how many BIOS have you developed to conclude it will "only" take 10 minutes.
laurencedeclan
Fanfold Paper
Posts: 5
Registered: ‎03-01-2019
Location: AU
Views: 607
Message 147 of 200

Re: ThinkPad E485/E585 - Firmware bug ACPI IVRS table

It is no longer necessary to use

iommu=pt

with the IVRS overrides on the latest 4.20.x kernels, unless you want kfd.

000000
Punch Card
Posts: 24
Registered: ‎01-03-2019
Location: AU
Views: 601
Message 148 of 200

Re: ThinkPad E485/E585 - Firmware bug ACPI IVRS table

Can that replace iommu=soft?

laurencedeclan
Fanfold Paper
Posts: 5
Registered: ‎03-01-2019
Location: AU
Views: 570
Message 149 of 200

Re: ThinkPad E485/E585 - Firmware bug ACPI IVRS table

If you are still using

iommu=soft

you should replace it with the two IVRS overrides for better performance. Take care that you are using the latest 4.20.x kernel.

MarcosRC
Fanfold Paper
Posts: 7
Registered: ‎12-18-2018
Location: AU
Views: 547
Message 150 of 200

Re: ThinkPad E485/E585 - Firmware bug ACPI IVRS table

It was mentioned by people that have done it (the ten minutes). Is just a data table where they have to enter the correct number.

 

When developing the next version of the BIOS, correcting that table won't add more than ten minutes.

Check out current deals!


Shop current deals

Top Kudoed Authors