Hi Lennart,
Do you have any idea that UBIFS supports writable memory mappings or not?
Regards,
Karnik Jain
On Wed, May 22, 2019 at 12:05 AM Kay One <kayone007@xxxxxxxxx> wrote:
Thank you so much for your time.Let me try changing JFFS2 to UBIFS as another available option to use for my SOC.I would keep you guys posted.Regards,Karnik JainOn Tue, May 21, 2019 at 8:18 PM Lennart Poettering <lennart@xxxxxxxxxxxxxx> wrote:On Di, 21.05.19 17:37, Kay One (kayone007@xxxxxxxxx) wrote:
> Thanks for the reply Lennart.
>
> 1) mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_SHARED, 28, 0) = -1 EINVAL
> > (Invalid argument) is the culprit for this error???
Yes.
> 2) If YES. Is this JFFS2 MTD file system bug or
It's a missing feature of the fs implementation, that the journal
relies on. Memory mapping a file is not a particularly exotic feature,
and yes, file systems really should implement that.
So my recommendation would be: work with the jffs2 folks to add this
to the kernel. They already have it for PROT_READ afaik, doing it for
PROT_WRITE is extra work, and not trivial but it will fix jffs2 for
every program that uses mmap for writing, and journald is just one of
them.
> 3) Is this not considered as systemd bug to support compatibility for such
> file system? Check first FS support writable mappings or not and if not
> then use other method to enable Persistence logging for systemd?
I think it's better to fix jffs2 in this regard than to add a
userspace workaround to every single consumer of it.
Doing this without mmap() is not trivial, and I don't see anyone
stepping up to do the work. It wouldn't precisely make our codebase
easier to read either...
Lennart
--
Lennart Poettering, Berlin
_______________________________________________ systemd-devel mailing list systemd-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/systemd-devel