I kept making 71MB initrd.gz but failing, though my little 1.44mb one worked

i had 2 methods of making (cpio and losetup). both seemed to read off usb and be ignored

i forgot to put both linuxrc (and optionally directory /initrd) in / of image before compressing it with gzip

i then got different error:

???? 00000

-------------------------

flew by my eyes upon boot, then I got

RAMDISK: incomplete write (4740 != 16377) ...

(follow by some ext2 read boundary errors and a message saying init was not found - though i'm not attempting an initramfs)

finally i decided my image couldn't be corrupt and time to try kernel parameters:

ramdisk_size=300000 # did the trick, yes it's for iniitrd not just initramfs

the (not so) funny thing is i read like 20 linux Q&A on other forums like stackexchange that danced around the question blaming the OS or the person's files. ugh!

Might try it with qemu loading it or place your stuff on a different vm or real hard drive to test if the usb is the issue.

When you boot to bios select the usb in the hard drive order as first in order and it should boot to sda I'd think.

emulation is slow. if i leave it on the "old PC" it will be faster than emulating on a "new chip"

ouch. it appears none of these work as module (note usb-storage compiled into kernel, a possible mistake i'm following up on)

ehci-hcd ohci-hcd uhci-hcd

it could be the "newish PC" having Intel Z170 chipset is not supported by 2.6.38 kernel modules (or, see above)

i'll just have to keep hacking a bit more to pin down if it's still a module order issue or if it's required upgrade for hw support or if new kernel even has support

----------------------------------

anyone out there ever hear of USB support for Intel Z170 boards with linux kernel being an issue ?

SOLVED. xhci-hcd was not tried and was the answer.

i was un-aware xhci-hcd even existed and that when tried the grub2 disk worked. sorry i asked the question to begin with.

to find this. first i compiled all the usb stuff as modules (for testing). during that i was preparing lk 4.x and happened to see xhci-hcd, typed it in thinking it wasn't in 2.6 about to give up and whack! it was there and loaded.

i had booted with a 71mb compressed loopbck ext2 filesystem (which i just cp'ed essential directories into, unmounted, then gzip'ed and named initrd.gz) having linuxrc script in / that only ran 'sh'. i found by surprise if root=/dev/ram0 that the script /linuxrc is skipped (used initrd.gz as real root), but when root=otherwise, the /linuxrc runs (and tries to mount root after exit).

I ALSO FOUND THAT COMPILING XHCI-HCD was helpful - INITRD IS THE HARDER WAY for anyone who finds compiling the linux kernel easily

in initrd i could load the modules and even mount the root: but upon exit the kernel claimed none of that existed (there are ways around this to tell kernel where root is so it isn't lost on initrd exit). when compiled into kernel it "just booted" into the ext2 partition of the flash drive simply.

SOLVED

furthermore i read more about QEMU and WINDOWS10 / HIGH SIERRA and KVM

i was wrong about "true virtualization" i though that was a thing of a dual boot computer running two OSes side by side (i had read that once long ago). i had no idea it was really "an accelerated emulator method", that one OS was "being emulated" and the other really running raw.

i had thought qemu (with accelerations) was an emulator and kvm was for running multiple kernels. i didn't realize it was integrated and leveraged to run under win10 and imac to run "almost at normal speeds"

So my fault on that. QEMU i a good consideration and sometime i will try that as a good or better alternative to having a bootable flash that is ultimately "keyed to a brand of PC".

yet i don't think i'd want to be stuck with either as my only way to run linux. either could end up becoming incompatible or inaccessible much later ?

thank you all for your advice

Thanks for the update and solution.