Hi Cameron,
I just submitted a patch for non-4Kb page size, please try it on POWER9.
Best Regards,
Guoju
On 2018/10/12 14:30, Cameron Berkenpas wrote:
Thanks!
Will this allow bcache to work with non-4k page sizes? Or just prevent
them from being corrupted?
As it turns out, 4k pages is the direction I want to go in anyway, but
it would be good to avoid corruption in case the system is
accidentally booted with 64k pages again.
On 10/11/2018 11:28 PM, guoju fang wrote:
It looks like the same problem that superblock will be corrupted when
PAGE_SIZE is not 4KB. I will submit a patch today.
On 2018/10/12 14:21, Cameron Berkenpas wrote:
Turns out the default page size is 64k on POWER!
Unfortunately, switching for me is a bit difficult right now...
Because I'm using BTRFS. BTRFS formatted for 64k page sizes won't
mount on a system configured for 4k page sizes...
I'll be able to try it out this weekend though and report back.
Thanks!
-Cameron
On 10/11/2018 08:07 PM, guoju fang wrote:
I've had a similar problem that superblock is corrupted when
PAGE_SIZE is 8KB.
On 2018/10/12 5:23, Cameron Berkenpas wrote:
It's probably important to mention that I'm currently running
kernel 4.18.13.
On 10/11/2018 02:18 PM, Cameron Berkenpas wrote:
Hello,
On my POWER9 machine running Debian ppc64EL (little endian), my
caching and backing devices mostly aren't recognized between
reboots.
The caching device is NEVER recognized between reboots.
The backing device will be recognized between reboots and the
filesystem will be accessible/mountable provided I have never
attached a caching device to it.
It doesn't matter if I detach the caching device before I reboot,
the backing device will not be recognized between reboots if it
has ever had a caching device attached.
The caching device after a reboot:
bcache-super-show /dev/sdxy
Invalid superblock (bad magic)
-bash: echo: write error: Invalid argument
If the caching device had been attached to the backing device,
the result is the same for the backing device.
-Cameron
