You can see some of the characters are corrupted - there are three
columns of this corruption down the screen, flickering like there's
some data flowing through the pixels in those characters, even when
it sits there waiting for the root file system.
Sounds like the kernel is using what counted for framebuffer
memory as
disk buffers now.
Ouch. While I fear this may be more than I'm capable of fixing, what
might be the best solution? My abilities stretch to compiling a
Looking at the screenshot, it does not look like disk buffers in the
screen memory - there is only minor corruption, and it rather looks
like
the console framebuffer driver overwrote certain characters the
wrong way
when scrolling the screen.
To verify this, you can let some script write full lines of varying
content to stdout (all 'a', all 'b' etc.) and look at the result. The
bottom row should always be OK, and artifacts show up in the other
rows.
Another test: compile a kernel with nothing but the bare minimum of
drivers (no ethernet, no disk, just ramdisk and keyboard) to see if
that
also shows screen corruption.
Michael