Arch Linux

Mr.I

Member

From: Kerbal

Registered: 2020-09-08

Posts: 2

extra/xrt installation issue

I tried to install the xrt package and I got some error messages. The system is up-to-date. I ran pacman -Syu and reboot before installing xrt.

Here's the error message:

$ sudo pacman -S xrt
warning: xrt-1:2.21.75-2 is up to date -- reinstalling
resolving dependencies...
looking for conflicting packages...

Packages (1) xrt-1:2.21.75-2

Total Installed Size:  47.63 MiB
Net Upgrade Size:       0.00 MiB

:: Proceed with installation? [Y/n] 
(1/1) checking keys in keyring                                              [###########################################] 100%
(1/1) checking package integrity                                            [###########################################] 100%
(1/1) loading package files                                                 [###########################################] 100%
(1/1) checking for file conflicts                                           [###########################################] 100%
(1/1) checking available disk space                                         [###########################################] 100%
:: Running pre-transaction hooks...
(1/1) Remove upgraded DKMS modules
==> dkms remove --no-depmod xrt-aws/2.21.0 -k 6.19.6-arch1-1
==> dkms remove --no-depmod xrt/2.21.0 -k 6.19.6-arch1-1
:: Processing package changes...
(1/1) reinstalling xrt                                                      [###########################################] 100%
:: Running post-transaction hooks...
(1/3) Arming ConditionNeedsUpdate...
(2/3) Install DKMS modules
==> dkms install --no-depmod xrt/2.21.0 -k 6.19.6-arch1-1
Deprecated feature: CLEAN (/usr/src/xrt-2.21.0/dkms.conf)

Error! Build of driver/xocl/userpf/xocl.ko failed for: 6.19.6-arch1-1 (x86_64)
Make sure the name and location of the generated module are correct,
or consult /var/lib/dkms/xrt/2.21.0/build/make.log for more information.
==> WARNING: `dkms install --no-depmod xrt/2.21.0 -k 6.19.6-arch1-1' exited 7
==> dkms install --no-depmod xrt-aws/2.21.0 -k 6.19.6-arch1-1
Deprecated feature: CLEAN (/usr/src/xrt-aws-2.21.0/dkms.conf)
==> depmod 6.19.6-arch1-1
(3/3) Updating linux initcpios...
==> Building image from preset: /etc/mkinitcpio.d/linux.preset: 'default'
==> Using default configuration file: '/etc/mkinitcpio.conf'
  -> -k /boot/vmlinuz-linux -g /boot/initramfs-linux.img
==> Starting build: '6.19.6-arch1-1'
  -> Running build hook: [base]
  -> Running build hook: [systemd]
  -> Running build hook: [autodetect]
  -> Running build hook: [microcode]
  -> Running build hook: [modconf]
  -> Running build hook: [kms]
  -> Running build hook: [keyboard]
  -> Running build hook: [keymap]
  -> Running build hook: [sd-vconsole]
  -> Running build hook: [block]
  -> Running build hook: [filesystems]
  -> Running build hook: [fsck]
==> Generating module dependencies
==> Creating zstd-compressed initcpio image: '/boot/initramfs-linux.img'
  -> Early uncompressed CPIO image generation successful
==> Initcpio image generation successful

So I followed the message and checked /var/lib/dkms/xrt/2.21.0/build/make.log, here's what I found:

xocl_drv.c: In function ‘xocl_cma_mem_alloc_huge_page’:
xocl_drv.c:1419:17: error: implicit declaration of function ‘DRM_ERROR’; did you mean ‘SO_ERROR’? [-Wimplicit-function-declaration]
 1419 |                 DRM_ERROR("Request %lld, round down to power of 2 %lld\n",
      |                 ^~~~~~~~~
      |                 SO_ERROR
xocl_drv.c: In function ‘xocl_cma_mem_alloc’:
xocl_drv.c:1599:17: error: implicit declaration of function ‘DRM_WARN’; did you mean ‘VM_WARN’? [-Wimplicit-function-declaration]
 1599 |                 DRM_WARN("Unable to allocate with page size 0x%llx", page_sz);
      |                 ^~~~~~~~
      |                 VM_WARN
xocl_drv.c: In function ‘xocl_cma_bank_alloc’:
xocl_drv.c:1672:25: error: implicit declaration of function ‘DRM_INFO’; did you mean ‘DRM_IO’? [-Wimplicit-function-declaration]
 1672 |                         DRM_INFO("HOST MEM already allocated, skip");
      |                         ^~~~~~~~
      |                         DRM_IO
make[5]: *** [/usr/lib/modules/6.19.6-arch1-1/build/scripts/Makefile.build:289: xocl_drv.o] Error 1
make[4]: *** [/usr/lib/modules/6.19.6-arch1-1/build/Makefile:2064: .] Error 2
make[3]: *** [/usr/lib/modules/6.19.6-arch1-1/build/Makefile:248: __sub-make] Error 2
make[3]: Leaving directory '/var/lib/dkms/xrt/2.21.0/build/driver/xocl/userpf'
make[2]: *** [Makefile:248: __sub-make] Error 2
make[2]: Leaving directory '/usr/lib/modules/6.19.6-arch1-1/build'
make[1]: *** [Makefile:137: all] Error 2
make[1]: Leaving directory '/var/lib/dkms/xrt/2.21.0/build/driver/xocl/userpf'
make: *** [Makefile:2: all] Error 2

Strange... So I checked my linux-headers and found that the missing symbols are defined in /usr/lib/modules/6.19.6-arch1-1/build/include/drm/drm_print.h.
So, the headers should be good.

Finally, I followed the build log and checked the faulty code file: /usr/src/xrt-2.21.0/driver/xocl/xocl_drv.h and /usr/src/xrt-2.21.0/driver/xocl/userpf/xocl_drv.c.
Guess what I found, it didn't include drm/drm_print.h at all!

So, I manually added a line in xocl_drv.h:

#include <drm/drm_print.h>

and then, dkms install again:

$ sudo dkms install --no-depmod xrt/2.21.0 -k 6.19.6-arch1-1
Deprecated feature: CLEAN (/var/lib/dkms/xrt/2.21.0/source/dkms.conf)
Sign command: /usr/lib/modules/6.19.6-arch1-1/build/scripts/sign-file
Signing key: /var/lib/dkms/mok.key
Public certificate (MOK): /var/lib/dkms/mok.pub

Building module(s)...................... done.
Signing module /var/lib/dkms/xrt/2.21.0/build/driver/xocl/userpf/xocl.ko
Signing module /var/lib/dkms/xrt/2.21.0/build/driver/xocl/mgmtpf/xclmgmt.ko
Installing /usr/lib/modules/6.19.6-arch1-1/updates/dkms/xocl.ko.zst
Installing /usr/lib/modules/6.19.6-arch1-1/updates/dkms/xclmgmt.ko.zst
Running depmod... done.

And success! It compiles.

I'm not sure what to do now. I'm not even sure if I did it right.
Because I think that xrt's developers won't make such elementary mistake. There should be something else went wrong.

So, I'm here to ask:

1. Did I do to the right thing and fixed a bug or did I miss something that actually caused this issue?
2. Should I submit a bug report?
3. If a bug report is needed, then to whom should it be submitted. (xrt developers or Arch Linux package maintainers?)

Scimmia

Fellow

Registered: 2012-09-01

Posts: 13,620

Re: extra/xrt installation issue

Does it build as-is with the -lts kernel? If something in the kernel headers previously pulled in that header but doesn't now, this is exactly what you'd see. If that's the issue, it should probably be reported upstream if it hasn't already been. Since the package isn't functional, though, reporting it to Arch with a link to the upstream issue makes sense. Side note, why does this package not depend on dkms?

Mr.I

Member

From: Kerbal

Registered: 2020-09-08

Posts: 2

Re: extra/xrt installation issue

Does it build as-is with the -lts kernel?

Yes, it does. I replaced linux and linux-headers with linux-lts and linux-lts-headers and the package can then be installed as is, without any modification.
But I can't just switch to linux-lts and call it a day though. I need the latest kernel, (waiting for) 7.0 to be specific.

Side note, why does this package not depend on dkms?

I also tried installing xrt with dkms uninstalled and all the dkms related hooks were simply skipped. No error message at all. Could this be the "missing piece" that caused those error messages?
I can't verify if xrt is working properly when installed this way because I'm not currently with the device that can use xrt. I'll post an update here later.

I think it won't work properly when installed without dkms. Because the xrt kernel modules are not built into the kernel, so dkms makes sense here.

--------------------------------------------------

I think I'll need to do more research before I submit a report. Thanks for the reply.

Last edited by Mr.I (Today 06:30:47)