unable to find a config file in /proc
arun54321 opened this issue · comments
On fedora config file is on /boot
>sudo booster -output /tmp/booster.img -strip -debug -force
unable to find a config file in /proc
Thanks for reporting it. Just to double-check, do you have any /proc/config* files at your system?
Could you please also provide output of ls /boot/?
ls /boot
booster.img refind
config-5.16.8-200.fc35.x86_64 refind_linux.conf
efi symvers-5.16.8-200.fc35.x86_64.gz
grub2 System.map-5.16.8-200.fc35.x86_64
initramfs-5.16.8-200.fc35.x86_64.img vmlinuz-5.16.8-200.fc35.x86_64
lost+found
ls /proc
1 1130 1257 1299 1404 1436 17 19 24 27 3217 3407 394 412 4253 4263 4328 46 53 69 bootconfig driver kcore meminfo self uptime
10 1155 1258 13 1405 1449 1745 1907 2401 2717 325 3411 3946 414 4254 4264 4329 4628 54 70 buddyinfo dynamic_debug keys misc slabinfo version
1029 1186 1260 1308 1406 1453 1763 2 2429 274 3256 35 395 415 4255 4265 4330 463 55 71 bus execdomains key-users modules softirqs vmallocinfo
1043 1190 1263 1331 1411 1455 1770 200 2483 28 326 36 4 418 4256 4268 4331 4635 6 805 cgroups fb kmsg mounts stat vmstat
1050 1191 1265 1347 1412 1461 1781 21 2551 29 3288 3671 40 4197 4257 4269 4385 464 609 81 cmdline filesystems kpagecgroup mtrr swaps zoneinfo
1063 12 1274 1348 1416 1479 1793 22 2553 3 3292 3673 402 42 4258 427 44 47 61 9 consoles fs kpagecount net sys
1072 1201 1279 1353 1425 15 18 222 2588 3000 33 37 403 424 4259 4287 445 48 63 94 cpuinfo interrupts kpageflags pagetypeinfo sysrq-trigger
1074 1203 1287 1369 1427 1521 1875 224 26 303 330 3731 406 425 426 43 45 49 64 95 crypto iomem latency_stats partitions sysvipc
11 1226 1293 1388 1428 1565 1879 23 261 31 331 38 407 4250 4260 431 4546 497 65 951 devices ioports loadavg pressure thread-self
1122 1238 1294 14 1432 16 1883 2310 2617 312 332 3808 41 4251 4261 4324 4547 504 66 acpi diskstats irq locks schedstat timer_list
1126 1253 1295 1402 1434 1684 1888 235 2653 32 34 39 410 4252 4262 4327 4555 524 68 asound dma kallsyms mdstat scsi tty
@arun54321 I added a change that implements this /boot/config-* feature. Please check wip branch and let me know if it makes any difference for you.
init - wip ❯ CGO_ENABLED=0 go build -trimpath -mod=readonly -modcacherw
go: github.com/anatol/clevis.go@v0.0.0-20211230012110-04c3cae00cc0 requires
github.com/anatol/tang.go@v0.0.0-20211230003118-4e3b011304ee requires
github.com/anatol/clevis.go@v0.0.0-20211227224505-77c31cb75e77 requires
github.com/anatol/tang.go@v0.0.0-20211208011427-4f66a321d8ef: invalid version: unknown revision 4f66a321d8ef
I don't know whether I'm doing something wrong, I get this error.
Oops sorry. I messed with dependencies earlier. I pushed a fix to wip branch. Please try it again.
go: github.com/anatol/clevis.go@v0.0.0-20220228201535-e211c818efc5 requires
github.com/anatol/tang.go@v0.0.0-20220228200145-7d0d81ac3cf3: missing go.sum entry; to add it:
go mod download github.com/anatol/tang.go
Hmm, it works fine for me:
➜ init git:(wip) go version
go version go1.17.7 linux/amd64
➜ init git:(wip) CGO_ENABLED=0 go build -trimpath -mod=readonly -modcacherw
go: downloading github.com/anatol/devmapper.go v0.0.0-20211210164347-f67e20c4e7f7
go: downloading github.com/anatol/smart.go v0.0.0-20220218195151-5ee9e8fa73f0
go: downloading github.com/anatol/luks.go v0.0.0-20211210165108-5d9a15b4f614
go: downloading github.com/anatol/clevis.go v0.0.0-20220228201535-e211c818efc5
go: downloading github.com/anatol/uevent.go v1.0.1-0.20210811163347-3e166d38c549
go: downloading github.com/google/go-tpm v0.3.3
go: downloading github.com/insomniacslk/dhcp v0.0.0-20220119180841-3c283ff8b7dd
go: downloading github.com/vishvananda/netlink v1.1.0
go: downloading github.com/yookoala/realpath v1.0.0
go: downloading golang.org/x/sys v0.0.0-20220227234510-4e6760a101f9
go: downloading gopkg.in/yaml.v3 v3.0.0-20210107192922-496545a6307b
go: downloading golang.org/x/net v0.0.0-20220225172249-27dd8689420f
go: downloading github.com/u-root/uio v0.0.0-20220204230159-dac05f7d2cb4
go: downloading golang.org/x/crypto v0.0.0-20220214200702-86341886e292
go: downloading github.com/vishvananda/netns v0.0.0-20211101163701-50045581ed74
go: downloading github.com/lestrrat-go/jwx v1.2.19
go: downloading github.com/pkg/errors v0.9.1
go: downloading github.com/lestrrat-go/iter v1.0.1
go: downloading github.com/lestrrat-go/backoff/v2 v2.0.8
go: downloading github.com/lestrrat-go/option v1.0.0
go: downloading github.com/lestrrat-go/blackmagic v1.0.0
go: downloading github.com/lestrrat-go/httpcc v1.0.0
What go version do you use?
go version go1.16.14 linux/amd64
I installed it from fedora repos.
I will try again after updating to 1.17.7
I believe clevis.go has been go mod tidy with 1.17 compat.
Does Fedora provide 1.17 go? If not then I will need to downgrade modules to 1.16.
@arun54321 let me know if you was able to get it compiled with 1.17.
Yes, it should work with 1.17+.
I don't see the error now. so I think it is working now.
https://0bin.net/paste/xPA+MbcG#kS4L3yChfWKg4VR-Tnl7ajrbCfOItbq7v7MGNEnqQ9I
But booster from this branch does not create initramfs image.
Booster fails for some reason:
enabling virtual console
detected language - '"en_US.UTF-8"'
2022/03/02 exit status 1
I need to look at the code to figure out why there is no any error message for this failure.
Also related question. Is golang 1.17 easily available for Fedore users? Is it ok to use 1.17 specific feature or it is going to be a lot of troubles for Fedora users?
My guess that there is a problem with loadkeys binary. I added a commit that prints more information in case of its failure. @arun54321 please pull the changes from wip and try it one more time. Share your debug output as well.
You can also disable vconsole option in your config file to skip this step.
exit status 1: loadkeys: Unable to open file: "us": No such file or directory
builds fine after disabling vconsole.
exit status 1: loadkeys: Unable to open file: "us": No such file or directory
thanks. What is the content of /etc/vconsole.conf and /etc/locale.conf?
what is output of loadkeys -q -b us?
Also related question. Is golang 1.17 easily available for Fedore users? Is it ok to use 1.17 specific feature or it is going to be a lot of troubles for Fedora users?
I'm using fedora for around a week or 10 days. I don't know about fedora release schedules. For a time being anyone could use my copr repo till it arrives on official repositories.
1.18 is coming in fedora 36 so It should not be problem in future.
/etc ❯ cat vconsole.conf
KEYMAP="us"
FONT="eurlatgr"
/etc ❯ cat locale.conf
LANG="en_US.UTF-8"
https://paste.centos.org/view/adb37d94
for loadkeys.
It looks like this in terminal.
If loadkeys -q -b us works for you then I don't really understand why booster with the same command.
Could you please post output of /etc/booster.yaml? And btw you can paste the text right here, without using screenshots.
compression: zstd
modules: hid_logitech_hidpp,i915,ext4
strip: true
universal: false
vconsole: falsewhen I paste output on github, they show few unknown symbols because I was using bat as cat in terminal.
Your configuration should call loadkeys -q -b us that runs successfully for you. So I do not really understand why it fails with Unable to open file: "us": No such file or directory.
Could you please reenable vconsole and then run booster with strace -f booster .... to get better understand what exactly booster is doing.
Looking at strace output I see that loadkeys called with correct parameters:
execve("/usr/bin/loadkeys", ["loadkeys", "-q", "-b", "\"us\""], 0xc000d05440 /* 68 vars */ <unfinished ...>
(which is loadkeys -q -b "us")
yet it fails with
[pid 12157] write(2, "loadkeys: ", 10) = 10
[pid 12157] write(2, "Unable to open file: \"us\": No su"..., 52 <unfinished ...>
[pid 12157] <... write resumed>) = 52
[pid 12157] write(2, "\n", 1 <unfinished ...>
[pid 12157] <... write resumed>) = 1
[pid 12157] exit_group(1) = ?
[pid 12157] +++ exited with 1 +++
[pid 12114] --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=12157, si_uid=1000, si_status=1, si_utime=0, si_stime=2} ---
[pid 12128] <... waitid resumed>{si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=12157, si_uid=1000, si_status=1, si_utime=0, si_stime=0}, WEXITED|WNOWAIT, NULL) = 0
[pid 12128] wait4(12157, <unfinished ...>
[pid 12128] <... wait4 resumed>[{WIFEXITED(s) && WEXITSTATUS(s) == 1}], 0, {ru_utime={tv_sec=0, tv_usec=2646}, ru_stime={tv_sec=0, tv_usec=27690}, ...}) = 12157
it is not clear for me why it happens. It might be that Fedora's loadkeys expects some env var or maybe something else.
Could you look at fedora's kbd patches?
https://kojipkgs.fedoraproject.org//packages/kbd/2.4.0/8.fc35/src/kbd-2.4.0-8.fc35.src.rpm
@arun54321 I actually found one more difference from my setup. The values in your /etc/vconsole.conf have quotes around values (KEYMAP="us"). I added code that strips these quotes. Hopefully it fixes the problem with loadkeys. Please take a look at wip branch.
booster - wip! ❯ g clean -f
Removing generator/booster
Removing init/init
booster - wip ❯ cd generator
generator - wip ❯ CFLAGS="-march=x86-64 -mtune=generic -O2 -pipe -fno-plt -fexceptions \
-Wp,-D_FORTIFY_SOURCE=2 -Wformat -Werror=format-security \
-fstack-clash-protection -fcf-protection"
generator - wip ❯ CXXFLAGS="$CFLAGS -Wp,-D_GLIBCXX_ASSERTIONS"
generator - wip ❯ LDFLAGS="-Wl,-O1,--sort-common,--as-needed,-z,relro,-z,now"
generator - wip ❯ CGO_CPPFLAGS="${CPPFLAGS}" CGO_CFLAGS="${CFLAGS}" CGO_CXXFLAGS="${CXXFLAGS}" CGO_LDFLAGS="${LDFLAGS}"
generator - wip ❯ go build -trimpath \
-buildmode=pie \
-mod=readonly \
-modcacherw \
-ldflags "-linkmode external -extldflags \"${LDFLAGS}\""
generator - wip! ❯ cd ../init
init - wip ❯ CGO_ENABLED=0 go build -trimpath -mod=readonly -modcacherw
init - wip! ❯ cd ..
booster - wip! ❯ sudo install -Dp -m755 generator/booster "/usr/bin/booster"
[sudo] password for arun:
booster - wip! ❯ sudo install -Dp -m755 init/init "/usr/lib/booster/init"
booster - wip! ❯ booster build /tmp/a.img -v
increasing open file limit 1024->524288
reading /boot/config-5.16.8-200.fc35.x86_64
no matches found for a device alias 'platform:efivars'
no matches found for a device alias 'acpi:LNXSYSTM:'
enabling virtual console
detected language - 'en_US.UTF-8'
2022/03/05 15:11:12 lstat /lib/libc.so.6: no such file or directory
booster - wip! ❯
libc file is in /usr/lib64. How do I compile / fix it properly?
I tried to symlink libc file to the destination.
Now I get open /usr/share/kbd/consolefonts/: no such file or directory
On fedora that folder is in /usr/lib/kbd/consolefonts
Image generation is working now after symlink those missing folders.
enabling virtual console
detected language - 'en_US.UTF-8'
font eurlatgr matched to file /usr/share/kbd/consolefonts/eurlatgr.psfu.gz
/ ❯ file /tmp/a.img
/tmp/a.img: Zstandard compressed data (v0.8+), Dictionary ID: None
libc file is in /usr/lib64. How do I compile / fix it properly?
I added /usr/lib64 path to the booster search list. Booster wip should work for you now without any extra symlinks.
On fedora that folder is in /usr/lib/kbd/consolefonts
I added a patch that makes it easier to configure this folder. Distros could update this path if needed.
051c8a1#diff-1dff658e5bba8c544f3d28aae04581865748e30fd711239342a747f63c4a406bR14
So the fedora package should patch booster path to /usr/lib/kbd/consolefonts
Image generation is working now after symlink those missing folders.
Could you please pull wip branch, remove all the symlinks you've made and then try booster one more time?
open /usr/share/kbd/consolefonts/: no such file or directory
The sourcecode needs to be patched to adjust it to Fedora's defaults.
Should I apply the patch or just pull and build again?
edit: ok
Works well.
Does it boot at Fedora?
Could you please also make a package for Fedora so other people can use the package easily? Is there any way to get this package to Fedora repository?
I don't know to write hooks to which rebuilds image during kernel updates.
Just now tested it. It boots to fedora.
Thank you.
Could you add force and kernelVersion flags to wip branch?
I am trying to write image generation script. It doesn't work due to missing flags.
So far, it works without those flags.
#!/bin/bash -e
# find out all installed kernels
kernels=($(rpm -q kernel-core))
for kernel in "$kernels"; do
version=$( echo $kernel | sed 's/kernel-core-//g' )
# write to tmp for testing
booster build /tmp/booster-${version}.img &
done
wait
It outputs
unknown flag `k'
unknown flag `f'
The flags start with double hype: --force. Please see man page for more information.
tmp ❯ booster --force build /tmp/booster.img
unknown flag `force'
--force and --kernel-version are build subcommand flags. i.e. you have to put it after the build codeword.
Here is an example of usage https://github.com/anatol/booster/blob/master/packaging/arch/regenerate_images
Thanks. Works fine now.
#!/bin/bash -e
# find out all installed kernels
kernels=($(rpm -q kernel-core))
for kernel in "$kernels"; do
version=$( echo $kernel | sed 's/kernel-core-//g' )
# write to tmp for testing
booster build --force --kernel-version ${version} /tmp/booster-${version}.img &
done
wait