Re: [PATCH] expand-on-first-boot: wait for udev to create symlink

["Schaffner, Tobias" <tobias.schaffner@siemens.com>]

On 09.12.22 01:22, Roberto A. Foglietta wrote:
> On Thu, 8 Dec 2022 at 18:39, Henning Schild <henning.schild@siemens.com> wrote:
>>
>> Am Thu, 8 Dec 2022 17:55:42 +0100
>> schrieb "T. Schaffner" <tobias.schaffner@siemens.com>:
>>
>>> From: Tobias Schaffner <tobias.schaffner@siemens.com>
>>>
>>> systemd-growfs depends on a symlink to the partition of the filesystem
>>> that should be resized. This symlink is created by udev in
>>> /dev/block/.
>>>
> 
> reading that patch I had a flash-back about two patches I did recently
> for another project
> 
> https://github.com/robang74/tinycore-editor/commit/d57271e3229551ddfbd3deba1363d2efe2914908
> https://github.com/robang74/tinycore-editor/commit/4ff435bb8a73c7e956ea4f9303a36fe06a136e54
> 
> both altered this file
> 
> https://github.com/robang74/tinycore-editor/blob/main/tinycore/changes/rcS
> 
> the main problem and the final result was much the same: trigger udev
> and wait for the dev link.
> 
> # Start Udev to populate /dev and handle hotplug events
> /sbin/udevd --daemon 2>&1 >/dev/null
> /sbin/udevadm trigger --action=add 2>&1 >/dev/null
> /sbin/udevadm settle --timeout=5
> echo -ne "\e[1;37m"
> for i in $(seq 1 10); do
>      if gettcdev; then
>      break;
> fi >/dev/null
> echo -n .
> sleep 1
> done
> echo -e "\e[0m"
> /sbin/udevadm control --exit --timeout=5
> 
> in my case, the booting device is not meaningful because the rootfs is a file.
> 
> function gettcdev() {
>      blkid --label $tclabel
> }
> 
> So, the search for the USB stick device is done s by label

Thanks for providing this Roberto!

In isar we have the systemd-udevd service which is responsible for the 
start of the udevd.

The problem appears if this systemd unit is not up yet when we try to 
use systemd-growfs by hand.

Therefore a udevadm trigger/settle would also most probably hit to early.

Correct me if I am wrong.

Best,
Tobias

>>
>>> /dev/block/ as +# systemd-growfs depends on it
>>> +while ! readlink /dev/block/* | grep -q "${LAST_PART_NAME}"; do
>>
>> I would make that much stricter to not be tricked by partial matches
>>
>> sda4 vs sda42
> 
> Irrelevant, if sda42 has been found then sda4 exists as well
> considering the problem is not yet started.
> However, for sake of precision and to avoid corner cases in which
> sda42 is created before sda4 a dollar at the end of regex will solve
> this issue.
> 
> +while ! readlink /dev/block/* | grep -qe "${LAST_PART_NAME}$"; do
> 
>>
>> readlink -f /dev/block/* | grep -e -q "^${LAST_PART}$"
>>
>>> +    sleep 0.1
>>
>> I wonder if there is anything we could do here. Maybe "udevadm trigger"
>> and depend on udev after all. Because the 5 is a nasty guess ...
> 
> This is the reason because I cited my patches. I hope they will help you.
> 
> Best regards, R-