Red Hat Bugzilla – Bug 52558
mem= kernel parameter forces root fs as ext2
Last modified: 2007-04-18 12:36:35 EDT
Specifying a mem=XXX parameter to an ext3-enabled kernel results in the
kernel mounting the root fs as ext2. This occurs only if XXX < actual RAM
I didn't see any odd errors in dmesg, and it appears jbd and ext3 loaded
successfully. I suspect this is a kernel bug, since 1.) I checked that
grub is passing the right parameters, and 2.) initrd is definitely trying
to mount / as ext3.
does dmesg say ANYTHING about initrd's ?
Tried it, tested it both on modular and non-modular ext3, and it works fine.
That was using LILO, though.
How do you know that initrd is trying to mount as ext3? If there's any log
message about that being attempted and failed, could you please add that to this
Hm, strange. It seems jbd and ext3 aren't being loaded before rc; in fact, it
seems initrd isn't being loaded at all. I don't get why, as grub is booting the
same config in both cases; I'm just adding 'mem=256M' by hand from the boot
screen. This may be a grub bug.
I've attached all pertinent logs.
Created attachment 29494 [details]
Created attachment 29495 [details]
Created attachment 29496 [details]
Created attachment 29497 [details]
Created attachment 29498 [details]
Created attachment 29499 [details]
Created attachment 29500 [details]
The attachments with a -mem prefix are the output when mem=256M is passed on the
command line. The others have no prefix, though in my testing mem=384M didn't
break anything (machine has 384MB of RAM.)
Created attachment 29501 [details]
My previous comment made no sense. Take two:
logs with -mem were the result of booting with mem=256M on the commandline.
logs with -nomem were plain boots with no mem= line. mem=384M doesn't seem to
Looks like grub loads the image outside of the memory the kernel can see
due to the mem=xxx.
That would be (sort of) of a grub bug
Still broken in 0.90-10
Fixed in 0.91-2 (if you want to use it on a 7.2-ish system, it should rebuild