In message [off-list ref] you wrote:
i suspect that leaving my root filesystem in flash would cause
noticeable performance problems (even if i managed to structure it so
What makes you think so?
The startup behaviour may change: at the moment you load the whole
root filesystem into RAM before you can run anything at all - that
means it takes longer before the first application can start. But
then everything _is_ in RAM, so anything else will start quickly,
too.
With a flash based root filesystem only the things (binaries,
libraries) that are actually needed for the first user processes
(init etc.) will be loaded to RAM, so they may actually start faster
than from a RAM disk. But anything else may require to load
additional libraries etc. so it may start slower than from a ramdisk.
On the other hand, some code or binaries (like error handling stuff,
debug tools etc.) may not be loaded at all for normal operation, so
you will have more RAM free, which adds to speed, too.
that it could be read-only). how hard would it be to automatically
have that root FS copied from JFFS2 to RAM and mounted from there?
This makes just no sense to me.
can that be done as a kernel option i haven't seen? or do i have to
do that manually, involving something like "pivot_root"? or is it
worth the trouble?
I cannot see any advantages from such a setup.
If you want a RAM based root filesystem, use a ramdisk.
Best regards,
Wolfgang Denk
--
Software Engineering: Embedded and Realtime Systems, Embedded Linux
Phone: (+49)-8142-4596-87 Fax: (+49)-8142-4596-88 Email: wd@denx.de
At the source of every error which is blamed on the computer you will
find at least two human errors, including the error of blaming it on
the computer.
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
On Mon, 28 Jun 2004, Wolfgang Denk wrote:
In message [off-list ref] you wrote:
quoted
that it could be read-only). how hard would it be to automatically
have that root FS copied from JFFS2 to RAM and mounted from there?
This makes just no sense to me.
ok, i'm probably getting confused about something fundamental here.
what i'm after is to have my root filesystem in persistent storage and
available for updates on the fly, hence the notion of partitioning off
a part of the system flash for the JFFS2 version of the root
filesystem. i've done that, so i've solved that problem -- i have a
root filesystem i can mount from a running system and modify.
now, i could just mount that JFFS2 version of the root fs straight out
of flash, but i get the impression that that would cause a performance
hit. so i was wondering about the feasibility about copying that
entire flash FS into RAM at boot time, and running out of RAM. this
would mean, of course, that if i wanted to make mods to the root FS,
i'd do them in flash and they'd only take effect at the next reboot,
which is fine with me.
or am i just hopelessly confused about what is and is not a reasonable
option here? i get the feeling i'm missing something important.
rday
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
On Mon, 2004-06-28 at 06:50 -0400, Robert P. J. Day wrote:
On Mon, 28 Jun 2004, Wolfgang Denk wrote:
quoted
In message [off-list ref] you wrote:
quoted
quoted
that it could be read-only). how hard would it be to automatically
have that root FS copied from JFFS2 to RAM and mounted from there?
This makes just no sense to me.
ok, i'm probably getting confused about something fundamental here.
what i'm after is to have my root filesystem in persistent storage and
available for updates on the fly, hence the notion of partitioning off
a part of the system flash for the JFFS2 version of the root
filesystem. i've done that, so i've solved that problem -- i have a
root filesystem i can mount from a running system and modify.
now, i could just mount that JFFS2 version of the root fs straight out
of flash, but i get the impression that that would cause a performance
hit.
Why so? JFFS2 is relatively slow to mount, and obviously _writing_ to
flash can be slow -- but for normal operation it should be fine. Try it.
Personally, I'd just use a JFFS2 root file system for everything. Put
your kernel in it too -- RedBoot and u-boot can load kernels from JFFS2,
and that means you can atomically update the kernel at runtime too,
which you can't do if it's stored in a separate partition (unless you
play with having two partitions for it and teach the bootloader to work
out which one is valid).
--
dwmw2
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
On Tue, 29 Jun 2004, David Woodhouse wrote:
On Mon, 2004-06-28 at 06:50 -0400, Robert P. J. Day wrote:
Why so? JFFS2 is relatively slow to mount, and obviously _writing_ to
flash can be slow -- but for normal operation it should be fine. Try it.
...
given that i have a few more questions (quelle surprise!) and that
they're more specifically related to filesystems and JFFS2, i figure i
might as well move this to the MTD mailing list. it's getting a bit
off-topic for a PPC list.
rday
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/