My 2.4.9 problem in cache_push

[Ken T. <ke...@we...>]

Hello,

I described yesterday how 53c7xx.c driver is hanging up during booting
after I turned on DEBUG in virgefb.c.

As best I can determine a cache_push call in 53c7xx.c is where the booting
stops.

The 53c7xx.c does __get_free_pages(GFP_ATOMIC,1); and gets 8192 bytes,
later calls cache_push() with the returned address and 8192 bytes.

During/after the the cache_push "Oops: kernel access of bad area, sig: 11"
appears in dmesgs.

I can 'fix' this two ways, either by getting more mem with
__get_free_pages(GFP_ATOMIC,3) which boots OK, or 
by commenting out the code following /* Also flush trailing block */ in
cache_push (arch/ppc/kernel/apus_setup.c), this moves the problem down to
the susequent call of cache_clear.  When the "trailing" code there is
commented out boot is again OK.

Is it OK to "flush trailing block" when the called with MAX_CACHE_SIZE ?
Or does it mean there is an address in the cache in the trailing block
that is invalid in that it's not mapped in ?

Ken.