On Mon, Mar 5, 2018 at 3:41 AM, <tang.junhui@xxxxxxxxxx> wrote:
>
> Hello Ryan Harper
>>
>> This happens in the initramfs for kernels which include bcache module. After the rootfs is
>> mounted and we've pivoted into system boot, then the udev cold-plug replay will repeat
>> the disk add/change events
> I run bcache in Redhat and Centos OS, I haven't met such issue.
> I am wondering, if udev cold-plug needs replay, it should also
> repeat the bcache change uevent which produced in bcache device
> registering, if so, the /dev/bcache/{by-uuid, by-label} should be
> also generated.
>>
>> but when the bcache register method runs a second time and
>> finds a backing device, a bcacheN block device already has it open, so the module exits out
>> the path of "device already registered" which prevents bcache from calling register_bdev()
>> which is the only path where a backing device could call bch_cached_dev_run() which is
>> responsible for emitting the KOBJ_CHANGE event which the bcache udev rules use to
>> generate the /dev/bcache/{by-uuid, by-label} events.
>
> I am just worry about that, If every disk change uevents triggerd a
> bcache device change uevents, there would be a lot of output messages
> like "/dev/bcache/{by-uuid, by-label} has existed, and it was
> created again..."
Udev does not generate messages when it creates symlinks as far as I know.
What you can see is the kernel messages when udev coldplugs devices:
[2681476.651631] bcache: register_bcache() error opening
/dev/nvme0n1p1: device already registered
[2681477.082824] bcache: register_bcache() error opening /dev/sda3:
device already registered
[3283536.544468] bcache: register_bcache() error opening
/dev/nvme0n1p1: device already registered
[3283537.018570] bcache: register_bcache() error opening /dev/sda3:
device already registered
[3567850.400367] bcache: register_bcache() error opening
/dev/nvme0n1p1: device already registered
[3567850.855402] bcache: register_bcache() error opening /dev/sda3:
device already registered
In general, I don't think there are a large number, or frequent bcache
CHANGE events outside of testing.
Ryan
>
> Thanks
> Tang Junhui
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
