Re: pre-2.1, AHA1542 fixes, WD7000

Eric Youngdale (
Thu, 26 Sep 1996 21:20:39 -0400

>One thing worries me, accessing the cdrom using workbone resulted in a
>zero buff address and a non-zero length, hence the "buff && SCSI_PA"
>around line 654. Eric, is this sensible? Also is the test "if
>(SCSI_PA(shpnt+1) > ISA_DMA_THRESHOLD)" around line 953 still needed since
>kmalloc supports DMA?

Both of these tests are really just sanity checks. The one at
953 was there because there was a time when kmalloc didn't have the GFP_DMA
flag. I guess the test could be removed, but if it were me I would probably
leave it just as another sanity check.

The zero buffer address is most likely caused by a command that
had no data transfer (i.e. TEST_UNIT_READY or something like it). Just
an offhand guess.