Debian-2.6.32 _LP64 compile issue

[lundman]

Hello, fresh Debian 6 install:

Linux huge 2.6.32-5-686 #1 SMP Wed Jan 12 04:01:41 UTC 2011 i686 GNU/Linux

spl-0.60rc1 compiles and insmod without issue.

zfs-0.60rc1 fails to compile due to:

/home/lundman/zfs/zfs-0.6.0-rc1/module/zfs/../../module/zfs/zfs_znode.c: In function ‘zfs_expldev’:
/home/lundman/zfs/zfs-0.6.0-rc1/module/zfs/../../module/zfs/zfs_znode.c:237: error: ‘NBITSMINOR32’ undeclared (first use in this function)
/home/lundman/zfs/zfs-0.6.0-rc1/module/zfs/../../module/zfs/zfs_znode.c:237: error: (Each undeclared identifier is reported only once
/home/lundman/zfs/zfs-0.6.0-rc1/module/zfs/../../module/zfs/zfs_znode.c:237: error: for each function it appears in.)
/home/lundman/zfs/zfs-0.6.0-rc1/module/zfs/../../module/zfs/zfs_znode.c:237: error: ‘MAXMAJ32’ undeclared (first use in this function)
/home/lundman/zfs/zfs-0.6.0-rc1/module/zfs/../../module/zfs/zfs_znode.c:239: error: ‘MAXMIN32’ undeclared (first use in this function)
make[7]: *** [/home/lundman/zfs/zfs-0.6.0-rc1/module/zfs/../../module/zfs/zfs_znode.o] Error 1

Attempted fix:

#define NBITSMINOR32    16
#define MAXMAJ32        0xffffUL
#define MAXMIN32        0xffffUL

But suspect it is wrong, since:

# ./cmd/zfs/zfs list
NAME        USED  AVAIL  REFER  MOUNTPOINT
zpool       150K  63.4M    21K  /zpool

# ls -la /zpool/

 kernel:[ 7261.068011] Oops: 0000 [#4] SMP 
Killed
 kernel:[ 7261.068011] EIP: [] generic_getxattr+0x25/0x5f SS:ESP 0068:c2d63e4c

Thanks for this work!

[behlendorf]

Your in luck, I just took care of these fixes earlier in the week. Grab the spl master branch, and the zfs linux-compat branch. These fixes are just getting a little more testing before being merged in to what will be tagged as 0.6.0-rc2. If you could give them a test on Debian that would be great.

https://github.com/behlendorf/spl/tarball/master

https://github.com/behlendorf/zfs/tarball/linux-compat

[lundman]

Sweet. I checked out master of spl, and linux-compat of zfs:

behlendorf-spl-5a52a78 
behlendorf-zfs-61da501

and both spl and zfs build without any issues at all. (excluding that I use --with-spl= to the spl build tree).

Alas, the ls bug is still present:

# ./cmd/zpool/zpool create -f zpool /var/tmp/disk1.bin 
root@huge:/home/lundman/zfs/behlendorf-zfs-61da501# ./cmd/zpool/zpool status  pool: zpool
 state: ONLINE
 scan: none requested
config:

        NAME                  STATE     READ WRITE CKSUM
        zpool                 ONLINE       0     0     0
          /var/tmp/disk1.bin  ONLINE       0     0     0


# ./cmd/zfs/zfs list
NAME    USED  AVAIL  REFER  MOUNTPOINT
zpool  85.5K  63.4M    21K  /zpool

# mkdir /zpool/roger
# rmdir /zpool/roger
# ./cmd/zfs/zfs create zpool/roger
# cp zfs.spec /zpool/
# cp /zpool/zfs.spec p
# diff p zfs.spec
# \ls /zpool
roger  zfs.spec

# \ls -l /zpool

Message from syslogd@huge at Feb 25 15:12:29 ...
 kernel:[  338.632017] Oops: 0000 [#4] SMP 
Killed
 kernel:[  338.632017] last sysfs file: /sys/devices/virtual/zpios/zpios/uevent
 kernel:[  338.632017] Process ls (pid: 1552, ti=f34de000 task=f6402640 task.ti=f34de000)
 kernel:[  338.632017] EIP: [] generic_getxattr+0x25/0x5f SS:ESP 0068:f34dfe4c

Still, very impressive. As long as I don't want to stat() the file, it is fully usable! :)

[behlendorf]

Great, I'm closing this bug for the build issue and I'll open a new one for your 'ls' crash. Based on the small bit you posted it looks like it's related to xattrs.