[Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

raphix · 2024-09-27 09:50:55

Since a few months ago, the internal keyboard of my laptop does not work after boot, if the kernel version is above 6.7.9 (excluded, this is the last working version I found).

My problem seems to be related to the one in this subject, but the touchpad problem seem to have been resolved by recent versions, but not mine.

My laptop is a HP Elitebook x360 1040 G8.

How can I check what the problem is, and/or find patches? (and how to apply them)

Last edited by raphix (2025-04-07 07:31:46)

gromit · 2024-09-27 10:00:49

I think the link to the other post you have mentioned is broken (it redirects me to the "create a new post" form) ...

seth · 2024-09-27 15:56:33

Also please post a complete system journal from a boot w/ a good and a bad kernel and, pure speculation, try to add "atkbd.reset i8042.reset i8042.nopnp i8042.kbdreset" to the https://wiki.archlinux.org/title/Kernel_parameters and see whether that makes any difference.

raphix · 2025-02-27 15:03:18

Hello,

First, thank you for your help. It didn't work, unfortunately, to add "atkbd.reset i8042.reset i8042.nopnp i8042.kbdreset" to my Grub starting line. However, I am not an expert in kernel parameters, perhaps it was not at the grub level that I should have tried. (I did check that it was used with `cat /proc/cmdline`, though).

I tried to re-install newer versions of the kernel (6.10.6, 6.10.10, 6.12.10, 6.13.2, 6.13.4), beyond the "working" 6.7.9), and discovered that there is an error message pointing towards the build of /var/lib/dkms/i8042-spectre/.
With 6.7.9, it's built without errors in /var/lib/dkms/i8042-spectre/kernel-6.7.9-arch1-1-x86_64/ (with a log in ./log/make.log). With all versions above, the build is in /var/lib/dkms/i8042-spectre/1.0/ (log in the ./build/make.log file).

Here's a log with errors:

DKMS (dkms-3.1.5) make.log for i8042-spectre/1.0 for kernel 6.10.6-arch1-1 (x86_64)
jeu. 27 févr. 2025 10:57:31 CET
Cleaning build area
# command: make -C /usr/lib/modules/6.10.6-arch1-1/build M=/var/lib/dkms/i8042-spectre/1.0/build clean
make : on entre dans le répertoire « /usr/lib/modules/6.10.6-arch1-1/build »
make : on quitte le répertoire « /usr/lib/modules/6.10.6-arch1-1/build »

# exit code: 0
# elapsed time: 00:00:00
Building module(s)
# command: make -j8 KERNELRELEASE=6.10.6-arch1-1 KDIR=/usr/lib/modules/6.10.6-arch1-1/build
make -C /usr/lib/modules/6.10.6-arch1-1/build M=/var/lib/dkms/i8042-spectre/1.0/build modules
make[1] : on entre dans le répertoire « /usr/lib/modules/6.10.6-arch1-1/build »
warning: the compiler differs from the one used to build the kernel
  The kernel was built by: gcc (GCC) 14.2.1 20240805
  You are using:           gcc (GCC) 14.2.1 20250207
  CC [M]  /var/lib/dkms/i8042-spectre/1.0/build/i8042-spectre.o
Dans le fichier inclus depuis /var/lib/dkms/i8042-spectre/1.0/build/i8042.h:14,
                 depuis /var/lib/dkms/i8042-spectre/1.0/build/i8042-spectre.c:131:
/var/lib/dkms/i8042-spectre/1.0/build/i8042-x86ia64io.h: Dans la fonction « i8042_pnp_id_to_string »:
/var/lib/dkms/i8042-spectre/1.0/build/i8042-x86ia64io.h:1004:9: erreur: déclaration implicite de la fonction « strlcpy »; vouliez-vous utiliser « strncpy » ? [-Wimplicit-function-declaration]
 1004 |         strlcpy(dst, "PNP:", dst_size);
      |         ^~~~~~~
      |         strncpy
make[3]: *** [scripts/Makefile.build:244: /var/lib/dkms/i8042-spectre/1.0/build/i8042-spectre.o] Error 1
make[2]: *** [/usr/lib/modules/6.10.6-arch1-1/build/Makefile:1934: /var/lib/dkms/i8042-spectre/1.0/build] Error 2
make[1]: *** [Makefile:240: __sub-make] Error 2
make[1] : on quitte le répertoire « /usr/lib/modules/6.10.6-arch1-1/build »
make: *** [Makefile:7: all] Error 2

# exit code: 2
# elapsed time: 00:00:01

As for the journal log that were asked for, here is the link for the 6.7.9 and the link for the 6.13.4.

I have just tried what is suggested in this post: https://bbs.archlinux.org/viewtopic.php?id=261108&p=3 (adding i8042.nopnp=1 i8042.dumbkbd=1 to kernal params), it did not wrok either.

It looks a bit like a typo in the source code (using strlcopy instead of strncopy), but:
- I don't know how to fix it (ie, manually download the code, change it, and compile it back)
- I am not sure if that would solve the problem

Last edited by raphix (2025-02-27 15:23:28)

seth · 2025-02-27 15:46:24

(I did check that it was used with `cat /proc/cmdline`, though).

Short of the actual module messages in dmesg this is authortive and the parameters can assumed to be applied.

/var/lib/dkms/i8042-spectre/1.0/build/i8042-x86ia64io.h: Dans la fonction « i8042_pnp_id_to_string »:
/var/lib/dkms/i8042-spectre/1.0/build/i8042-x86ia64io.h:1004:9: erreur: déclaration implicite de la fonction « strlcpy »; vouliez-vous utiliser « strncpy » ? [-Wimplicit-function-declaration]
 1004 |         strlcpy(dst, "PNP:", dst_size);
      |         ^~~~~~~
      |         strncpy
make[3]: *** [scripts/Makefile.build:244: /var/lib/dkms/i8042-spectre/1.0/build/i8042-spectre.o] Error 1
make[2]: *** [/usr/lib/modules/6.10.6-arch1-1/build/Makefile:1934: /var/lib/dkms/i8042-spectre/1.0/build] Error 2

The build fails, /var/lib/dkms/i8042-spectre/1.0/build/i8042-x86ia64io.h needs to "#include <string.h>"

https://github.com/dustydecapod/i8042-s … s/issues/4
https://github.com/dustydecapod/i8042-s … s/issues/5

raphix · 2025-03-02 10:16:49

Currently, no luck with this github repo either.

My problem seems a bit different: I cannot build the i8042 module for kernel versions beyond 6.7.9. So, the workaround in issue #5 is to help people successfully building i8042 but getting a wrong/unexpected behavior (keys unresponsive for 10-20 seconds at boot). And issue #4 has no answer, just a build error message.

seth · 2025-03-02 14:36:27

Edit /var/lib/dkms/i8042-spectre/1.0/build/i8042-x86ia64io.h and add "#include <string.h>" ?

raphix · 2025-04-03 07:29:21

I have tried to edit and add this line, but it still fails with C compilation errors (e.g., use of strlcopy or strncopy). I have tried to "fix" a few of them, but I am not a C/module expert, so I did not yet succeed in obtaining a clean build :-(

It seems that a recent update introduced some problems with kernel 6.7.9 and suspend/power management: I suspect that I will have to dive more into this problem, to be able to run newer kernels (which don't have "suspend" problems, but currently don't work with my laptop's internal keyboard).

seth · 2025-04-03 20:24:31

still fails with C compilation errors (e.g., use of strlcopy or strncopy)

Whenever it complains about the imlicit declaration of strlcopy, the file needs to include string.h

In doubt

LC_ALL=C make

and post the output.

raphix · 2025-04-04 08:29:46

First, thanks a lot for your time.

After trying a bit, I guess I don't precisely get where I have to put the #include <string.h> line.

Here is the folder structure in /var/lib/dkms/i8042-spectre/1.0/build:

.
├── dkms.conf
├── i8042.h
├── i8042-spectre.c
├── i8042-spectre-dkms
│   ├── dkms.conf
│   ├── i8042.h
│   ├── i8042-spectre.c
│   ├── i8042-x86ia64io.h
│   ├── Makefile
│   └── README.md
├── i8042-spectre.mod
├── i8042-x86ia64io.h
├── Makefile
├── make.log
└── README.md

I have tried adding the line as the first line (to avoid ifdef branching) of both files named "i8042-x86ia64io.h" (ie, i8042-x86ia64io.h and i8042-spectre-dkms/i8042-x86ia64io.h), but the compilation fails:

[build]# LC_ALL=C  make
make -C /lib/modules/6.13.8-arch1-1/build M=/var/lib/dkms/i8042-spectre/1.0/build modules
make[1]: Entering directory '/usr/lib/modules/6.13.8-arch1-1/build'
make[2] : on entre dans le répertoire « /var/lib/dkms/i8042-spectre/1.0/build »
  CC [M]  i8042-spectre.o
Dans le fichier inclus depuis i8042-spectre.c:131:
i8042.h:14:10: erreur fatale: string.h : Aucun fichier ou dossier de ce nom
   14 | #include <string.h>
      |          ^~~~~~~~~~
compilation terminée.
make[4]: *** [/usr/lib/modules/6.13.8-arch1-1/build/scripts/Makefile.build:196: i8042-spectre.o] Error 1
make[3]: *** [/usr/lib/modules/6.13.8-arch1-1/build/Makefile:1989: .] Error 2
make[2]: *** [/usr/lib/modules/6.13.8-arch1-1/build/Makefile:251: __sub-make] Error 2
make[2] : on quitte le répertoire « /var/lib/dkms/i8042-spectre/1.0/build »
make[1]: *** [Makefile:251: __sub-make] Error 2
make[1]: Leaving directory '/usr/lib/modules/6.13.8-arch1-1/build'
make: *** [Makefile:7: all] Error 2

I have also tried to edit and then compile from /var/lib/dkms/i8042-spectre/1.0/source, with the same results. I have tried with #include <linux/string.h>, but then it complains about "implicit declaration of strlcpy" (see full message at the top of this topic).

So could you point me to a documentation or explain me a bit where I should add the line and how to compile?
(thank you again for your time!)

Last edited by raphix (2025-04-04 08:30:14)

seth · 2025-04-04 15:04:51

Why does LC_ALL=C make speak french?

Either way, /usr/include/string.h is part of glibc

pacman -Qikk glibc
stat /usr/include/string.h
VERBOSE=1 LC_ALL=C make

and try to add the include to i8042-spectre.c

raphix · 2025-04-04 20:23:53

For the locale thing, I don't know what's going on… I even tried with en_EN.UTF-8, same result (French messages). I had error messages in English only when using the "unset LANG; source /etc/profile.d/locale.sh" (and creating a ".config/locale.conf" file, with "LANG=en_US.UTF-8" to override the system-wide setting).

I tried to add the line to i8042-spectre.c, still no result.

seth · 2025-04-05 06:54:07

Does the VERBOSE=1 at least increase the compiler output level so we can see the include paths etc.?
What about the file stat and glibc sanity?

(The french output is more of a curiosity as it might indicate a context isolation to explain the missing header - I can still read that )

raphix · 2025-04-06 09:02:40

$ sudo pacman -Qikk glibc | curl -F 'file=@-' 0x0.st
http://0x0.st/8_c2.txt

$ stat /usr/include/string.h | curl -F 'file=@-' 0x0.st
http://0x0.st/8_c9.txt

As for the VERBOSE=1, the output are the same with and without it.

Yes, the French messages are unexpected, and perhaps it's a lead towards an anwser (but I haven't found yet how)…

Well, I don't exactly understand how "modules" are handled but: it seems that this i8042-spectre is broken, and has been for years (as evidenced by the github repo posted above by Seth). However, it is widely used by a lot of laptops (HP for business), not a niche product. Isn't there a "replacement" module that I can try?

Last edited by raphix (2025-04-06 09:31:06)

seth · 2025-04-06 14:36:54

https://github.com/archlinux/linux/comm … f8b85e6b44
The module doesn't tap into the userspace glibc, so…

sed -i 's/strlcpy/strscpy/g' i8042-spectre.c i8042-x86ia64io.h
make

raphix · 2025-04-06 19:11:34

Well, thank you very much! I thought I had checked that, but this helped a lot.
There is an error after that, which I was able to correct (a function returning no longer returning an int).
Now, the "make" goes through!

However, I need one extra step: when I use "pacman -S linux linux-headers", the folder with those updates seems to be ignored, and I get the same error. The installation process uses "dkms install --no-depmod i8042-spectre/1.0 -k 6.13.8-arch1-1".

Sorry for being a noob, but what should I do after the "make" is OK?

Last edited by raphix (2025-04-06 19:12:22)

seth · 2025-04-06 19:34:13

There is an error after that, which I was able to correct (a function returning no longer returning an int).

Specifically?
(The return values of strlcpy and strscpy differ in nature, I didn't see them being used but glossing over that might result in a building but misbehaving module!)

The dkms sources will reside in /usr/src, you'd have to patch them there, too.

raphix · 2025-04-07 07:27:53

First, thank you very much for all this, it works now! (after patching in "/usr/src")
I have a recent kernel (so, I can suspend my laptop) and a keyboard!

You're right, I should have posted the log and the second error.

In i8042-spectre.c, line 1603 and following:

static int i8042_remove(struct platform_device *dev)
{
        i8042_unregister_ports();
        i8042_free_irqs();
        i8042_controller_reset(false);
        i8042_platform_device = NULL;

        return 0;
}

static struct platform_driver i8042_driver = {
        .driver         = {
                .name   = "i8042",
#ifdef CONFIG_PM
                .pm     = &i8042_pm_ops,
#endif
        },
        .remove         = i8042_remove,
        .shutdown       = i8042_shutdown,
};

The error message is the following (sorry, still French here ):

make -C /lib/modules/6.13.8-arch1-1/build M=/usr/src/i8042-spectre-1.0/i8042-spectre-dkms modules
make[1] : on entre dans le répertoire « /usr/lib/modules/6.13.8-arch1-1/build »
make[2] : on entre dans le répertoire « /usr/src/i8042-spectre-1.0/i8042-spectre-dkms »
  CC [M]  i8042-spectre.o
Dans le fichier inclus depuis i8042.h:14,
                 depuis i8042-spectre.c:131:
i8042-x86ia64io.h: Dans la fonction « i8042_pnp_id_to_string »:
i8042-x86ia64io.h:1005:9: erreur: déclaration implicite de la fonction « strlcpy »; vouliez-vous utiliser « strncpy » ? [-Wimplicit-function-declaration]
 1005 |         strlcpy(dst, "PNP:", dst_size);
      |         ^~~~~~~
      |         strncpy
i8042-spectre.c: Au plus haut niveau:
i8042-spectre.c:1620:27: erreur: initialisation de « void (*)(struct platform_device *) » depuis le type pointeur « int (*)(struct platform_device *) » qui est incompatible [-Wincompatible-pointer-types]
 1620 |         .remove         = i8042_remove,
      |                           ^~~~~~~~~~~~
i8042-spectre.c:1620:27: note: (près de l'initialisation de « i8042_driver.remove »)
make[4]: *** [/usr/lib/modules/6.13.8-arch1-1/build/scripts/Makefile.build:196: i8042-spectre.o] Error 1
make[3]: *** [/usr/lib/modules/6.13.8-arch1-1/build/Makefile:1989: .] Error 2
make[2]: *** [/usr/lib/modules/6.13.8-arch1-1/build/Makefile:251: __sub-make] Error 2
make[2] : on quitte le répertoire « /usr/src/i8042-spectre-1.0/i8042-spectre-dkms »
make[1]: *** [Makefile:251: __sub-make] Error 2
make[1] : on quitte le répertoire « /usr/lib/modules/6.13.8-arch1-1/build »
make: *** [Makefile:7: all] Error 2

So my fix was just to change the "static int i8042_remove(struct platform_device *dev)" (line 1603) into "static void i8042_remove(struct platform_device *dev)", and to remove the last two lines of the function ("return 0"), so that it does not return anything explictly.

Perhaps I should have done the converse, i.e., find the declaration and change it to "int", instead of changing the function to respect the declaration.

Last edited by raphix (2025-04-07 07:30:33)

seth · 2025-04-07 13:50:14

Nope, void* is fine, see https://elixir.bootlin.com/linux/v6.13. … ice.h#L236

Arch Linux

#1 2024-09-27 09:50:55

[Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#2 2024-09-27 10:00:49

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#3 2024-09-27 15:56:33

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#4 2025-02-27 15:03:18

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#5 2025-02-27 15:46:24

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#6 2025-03-02 10:16:49

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#7 2025-03-02 14:36:27

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#8 2025-04-03 07:29:21

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#9 2025-04-03 20:24:31

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#10 2025-04-04 08:29:46

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#11 2025-04-04 15:04:51

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#12 2025-04-04 20:23:53

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#13 2025-04-05 06:54:07

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#14 2025-04-06 09:02:40

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#15 2025-04-06 14:36:54

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#16 2025-04-06 19:11:34

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#17 2025-04-06 19:34:13

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#18 2025-04-07 07:27:53

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

#19 2025-04-07 13:50:14

Re: [Fixed] (i8042-spectre) laptop keyboard disabled in recent kernels

Board footer