kernel panic, Areca, 6.2-REL with GENERIC [Was: Re: 6.2-st panic w/ ggate [Was: Re: kernel panic on 6.2-RC2 with GENERIC.]]

[Isaac Levy]

On Aug 29, 2007, at 9:04 AM, Scott Long wrote:

> Isaac Levy wrote:
>> Hi All,
>> I hate to say, I'm picking up from this thread:
>> http://lists.freebsd.org/pipermail/freebsd-stable/2007-January/ 
>> 031970.html
>> On Feb 15, 2007, at 11:06 AM, Scott Long wrote:
>>> I don't see how they could be related.  I've committed a fix for  
>>> the Areca problem.
>>>
>>> Scott
>> I've reproduced the Areca problem from previous threads this year,  
>> and was wondering if it's still broken, or if there's anything I'm  
>> missing?
>> Here's what happens:
>> <snip>
>> Aug 29 01:56:38 greengoose sudo:      ike : TTY=ttyp6 ; PWD=/usr/ 
>> home/ike ; USER=root ; COMMAND=/usr/sbin/periodic daily
>> Aug 29 01:57:29 greengoose kernel: g_vfs_done():da2p1[WRITE 
>> (offset=5998165426176, length=131072)]error = 5
>> Aug 29 01:57:58 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=27556233216, length=16384)]error = 5
>> Aug 29 01:57:58 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=27556249600, length=16384)]error = 5
>> Aug 29 01:57:58 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=27556216832, length=16384)]error = 5
>> Aug 29 01:57:58 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=27556298752, length=16384)]error = 5
>> Aug 29 01:57:58 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=27556315136, length=16384)]error = 5
>> Aug 29 01:57:58 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=27556282368, length=16384)]error = 5
>> Aug 29 01:57:58 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=27556331520, length=16384)]error = 5
>>   <repeat around 1710 times>
>> Aug 29 02:08:50 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=122359873536, length=16384)]error = 5
>> Aug 29 02:08:50 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=122359889920, length=16384)]error = 5
>> Aug 29 02:08:50 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=122359939072, length=16384)]error = 5
>> Aug 29 02:08:50 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=122359955456, length=16384)]error = 5
>> Aug 29 02:08:50 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=122360086528, length=16384)]error = 5
>> Aug 29 02:08:50 greengoose kernel: g_vfs_done():da0s1f[WRITE 
>> (offset=28133228544, length=16384)]error = 5
>>   <pop goes the kernel>
>> Aug 29 02:11:40 greengoose syslogd: kernel boot file is /boot/ 
>> kernel/kernel
>> <snip>
>> Here's the specs:
>> FreeBSD 6.2-REL, i386 GENERIC kernel (installed today [many times  
>> over])
>> da2: <Areca ARC-1160-VOL#02 R001> Fixed Direct Access SCSI-5 device
>>   (12 chhannel SATA Raid, Maxtor disks, with 'Disable Maxtor'  
>> under 'HDD Read-Ahead Cache' set)
>> Firmware Version - V1.43 2007-4-17
>> BOOT ROM Version - V1.41 2006-5-24
>> Additional info:
>>   - Volumes are Raid5 and Pass-Through drives
>>   - Pass-Through drives fail just like the Raid5 volumes do
>>   - Each time it crashes, disk i/o is SCREAMING'.
>> -- 
>> If I've missed something, please feel free to point me to the url!
>> But if not, it seems whatever bug in this driver existed, did not  
>> go away...
>> (Sidenote: Gotta give a big thanks to the whoever is working on  
>> these drivers, once you get working with big disks, it takes A  
>> LONG TIME to figure out what's wrong...  A fsck takes forever...   
>> Not whining, actually I'm really meaning to thank the driver  
>> writers for wading through this thick muck!!!)
>
> There are locking problems in the arcmsr driver.  I fixed them in  
> HEAD but not in 6-STABLE yet.  I really don't have time today to  
> generate patch; I encourage someone to look at the commit I made on  
> July 31 and
> backport those changes to 6-STABLE.
>
> Scott

That's excellent to hear!

For the record, are these the most current changes you speak of?:
(I'm reading through the source, perhaps eventually I'll be able to  
help more...)
http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/dev/arcmsr/

- I am not capable of backporting the changes,
+ but I will gladly test any changes this with a GENERIC kernel once  
they make it to STABLE

Rocket,
.ike