Bug 2700

Summary:	Kernel panic: Buffer at physical address > 16Mb used for aha1542
Product:	SCSI Drivers	Reporter:	Ralf Angeli (angeli)
Component:	Other	Assignee:	Mike Anderson (andmike)
Status:	REJECTED UNREPRODUCIBLE
Severity:	normal	CC:	bunk, protasnb, ralf
Priority:	P2
Hardware:	i386
OS:	Linux
Kernel Version:	2.6.6	Subsystem:
Regression:	Yes	Bisected commit-id:

Description Ralf Angeli 2004-05-13 09:34:31 UTC

Distribution: Debian/sid
Hardware Environment: PIII 700,
Software Environment: N/A
Problem Description: Kernel panics as soon as the module aha1542 is loaded.

Steps to reproduce:
1. Boot.
2. # modprobe aha1542

This is probably a dupe of bug 2390.  But it doesn't say anything about a
kernel panic, so this here might be a bit clearer.

Comment 1 Ralf Hoffmann 2005-03-07 02:56:03 UTC

I have the same kernel panic on my system. But the aha1542 module works fine
until I insmod the sr-mod module for cdrom support. The kernel panic appears at
this moment. The panic is called in drivers/scsi/aha1542.c:750

if (buff && SCSI_BUF_PA(buff + bufflen - 1) > ISA_DMA_THRESHOLD)
	BAD_DMA(buff, bufflen);

SCSI_BUF_PA is defined to isa_virt_to_bus(). In other kernel modules the value
of this function is not checked against valid dma sizes. Commenting this test
prevent the kernel panic and the scsi controller and cdrom drive work on my
system. I'm not sure if this test is really wrong or there is now some bad dma
transfer to wrong memory areas.

On the other hand in the 2.4 kernel this problem does not appear and the test
for incorrect dma buffer does not fail.

Comment 2 Natalie Protasevich 2007-09-03 23:02:13 UTC

Do you still have same problem with new kernels? (and still own the hardware)
Thanks.

Comment 3 Natalie Protasevich 2008-03-03 20:08:31 UTC

It's been a while since the bug was reported. Since no updates, I'm closing it.