WARNING! 3ware controllers break SUSPEND.
 

Hi,
After a week of trying to get my 3ware 9650se raid controller working on Suse 11.1, I finally got past the triage guy and contacted an engineer. He, Adam Radford, told me, and I quote:
"Yes, do not support 's2ram', that is the problem. When you wake up after suspend, we are
not sending an 'initconnection' to our firmware from the driver that tells the firmware
to re-detect SGL's as 64-bit and it is detecting them as 32-bit, that is why you see
'Illegal SGL length' errors.

I will try to implement 's2ram' support ASAP. You are the first customer to request
s2ram support for 3ware. Until then, don't try to suspend the machine if you have the 3ware
card in it. Just do a normal shutdown/bootup when you want to save power. ".

When you suspend and then wake up, the card stays in some type of zombie state. The OS sees the card freaking out and does an fsck, potentially destroying all of your data or at best keeping you waiting an hour before finishing the fsck.

Apparently, I am the only 3ware customer to ever wonder why suspend does not work.

I asked that they put a disclaimer on their web site warning potential buyers about this defect.

Let the buyer beware,

BrianP

Possibly because most RAID card users have them in servers or workstations that are either on or off. Servers virtually are never set to suspend and few workstations either. It's likely not going to be used very oftern but it would be good for it to be properly implemented.

SUSPEND fixed on Suse 11.1, 64 bit for 3Ware 9650se!
 

Hi,
One of the engineers at 3Ware (Adam) developed a new kernel module for the 3w-9xxx level controllers. After a few compiling problems, I now have `s2ram -f -a 3` working. It takes about a second to get to sleep and roughly 5 to wake up. What a relief.

I had to hack the Makefile to get it to find my kernel source directory:
godzilla2:~/temp/driver # diff Makefile Makefile.orig
3,4d2
< SRC := /usr/src/linux-2.6.27.7-9-obj/x86_64/default

I looked at my Ubuntu /usr/src/linux tree and it is quite different from the way Suse does it.

Then, just;
make clean
make
cp ./3w-9xxx.ko /lib/modules/2.6.27.7-9-default/kernel/drivers/scsi
mkinitrd
sync
reboot

One other difficulty was that YAST apparently lied to me about kernel-source and kernel-syms. It showed them both checked and `rpm -qi <module>` showed a page of output for each one. When I kept getting "no rule to make ..." errors I looked for Makefiles system wide and found that they were all on my Ubuntu drive.

Going into yast and updating kernel source and symbols fixed the problem.

Thanks to 3Ware for a fast fix for this problem!

BrianP

How to fix
 

Hi,

I have the same problem; 9650se with ubuntu 8.10.

Just a few questions about your fix: I would have to compile it into the kernel every thime there is a kernel update?

And do you have any information whether this will ever enter the regular kernel, so I don't have to worry about compiling?

Strangely suspend works with Windows XP. And would you mind sharing this kernel module with me?

Regards,
Mogliii

same here! please brian, let us know where to get the kernel module, it would be very much appreciated!

cheers, rooots.

Hi,

Same problem with a 3ware 8XXX dual channel (for workstations). Looks like there is no access to disks after a resume.

Please keep us posted.

Regards,

Please ask Adam to check this into the main driver. :) I'm running into this under Fedora Core 11 as well (3ware 9650).

- Chris