Commit Graph

5686 Commits

Author SHA1 Message Date
Brian Behlendorf 40a39e1103 Conserve stack in dsl_scan_visit()
The dsl_scan_visit() function is a little heavy weight taking 464
bytes on the stack.  This can be easily reduced for little cost by
moving zap_cursor_t and zap_attribute_t off the stack and on to the
heap.  After this change dsl_scan_visit() has been reduced in size
by 320 bytes.

This change was made to reduce stack usage in the dsl_scan_sync()
callpath which is recursive and has been observed to overflow the
stack.

Issue #174
2011-04-26 15:48:00 -07:00
Brian Behlendorf b81c4ac9af Conserve stack in dsl_scan_visitbp()
This function is called recursively so everything possible must be
done to limit its stack consumption.  The dprintf_bp() debugging
function adds 30 bytes of local variables to the function we cannot
afford.  By commenting out this debugging we save 30 bytes per
recursion and depths of 13 are not uncommon.  This yeilds a total
stack saving of 390 bytes on our 8k stack.

Issue #174
2011-04-26 15:48:00 -07:00
Brian Behlendorf 7a060636b0 Conserve stack in dsl_scan_visitbp()
The recursive call chain dsl_scan_visitbp() -> dsl_scan_recurse() ->
dsl_scan_visitdnode() -> dsl_scan_visitbp has been observed to consume
considerable stack resulting in a stack overflow (>8k).  The cleanest
way I see to fix this with minimal impact to the existing flow of
code, and with the fewest performance concerns, is to always inline
dsl_scan_recurse() and dsl_scan_visitdnode().  While this will increase
the function size of dsl_scan_visitbp(), by 4660 bytes, it also reduces
the stack requirements by removing the function call overhead.

Issue #174
2011-04-26 13:37:35 -07:00
Brian Behlendorf 44e9e34793 Merged pull request #212 from dajhorn/hostid.
Use gethostid in the Linux convention.
2011-04-26 13:30:27 -07:00
Brian Behlendorf 701b1f8168 Fix zvol deadlock
It's possible for a zvol_write thread to enter direct memory reclaim
while holding open a transaction group.  This results in the system
attempting to write out data to the disk to free memory.  Unfortunately,
this can't succeed because the the thread doing reclaim is holding open
the txg which must be closed to be synced to disk.  To prevent this
the offending allocation is marked KM_PUSHPAGE which will prevent it
from attempting writeback.

Closes #191
2011-04-26 12:56:35 -07:00
Brian Behlendorf 4c16d2471a Merged pull request #40 from dajhorn/spl-proc-typos.
Correct typos in the spl proc handler.
2011-04-25 14:51:48 -07:00
Darik Horn 492b8e9e7b Use gethostid in the Linux convention.
Disable the gethostid() override for Solaris behavior because Linux systems
implement the POSIX standard in a way that allows a negative result.

Mask the gethostid() result to the lower four bytes, like coreutils does in
/usr/bin/hostid, to prevent junk bits or sign-extension on systems that have an
eight byte long type. This can cause a spurious hostid mismatch that prevents
zpool import on 64-bit systems.
2011-04-25 10:36:17 -05:00
Darik Horn c95b308d12 Correct typos in the spl proc handler.
Correct a format typo that causes /proc/sys/kernel/spl/hostid
to return a decimal number instead of a hexadecimal number.
2011-04-24 20:56:07 -05:00
Brian Behlendorf a1cc0b3290 Fix 32-bit MAXOFFSET_T definition
Having MAXOFFSET_T defined to 0x7fffffffl was artificially limiting
the maximum file size on 32-bit systems.  In reality MAXOFFSET_T is
used when working with 'long long' types and as such we now define
it as LLONG_MAX.  This resolves the 2GB file size limit for files
and additionally allows zvols greater than 2GB on 32-bit systems.

Closes #136
Closes #81
2011-04-22 16:21:26 -07:00
Brian Behlendorf d837ae395b Fix 32-bit MAXOFFSET_T definition
The correct definition of MAXOFFSET_T under Solaris is in reality
tied to the maximum size of a 'long long' type.  With this in mind
MAXOFFSET_T is now defined as LLONG_MAX which ensures the correct
value is used on both 32-bit and 64-bit systems.
2011-04-22 16:17:13 -07:00
Brian Behlendorf e2448b0e62 Fix spurious -EFAULT when setting I/O scheduler
Occasionally we would see an -EFAULT returned when setting the
I/O scheduler on a vdev.  This was caused an improperly formatted
user mode helper command.

This commit restructures the command to something simpler, allocates
space for it dynamically to save stack, and removes the retry logic
which is no longer needed.

Closes #169
2011-04-22 14:55:35 -07:00
Brian Behlendorf 6a8f9b6bf0 Enforce ARC meta-data limits
This change ensures the ARC meta-data limits are enforced.  Without
this enforcement meta-data can grow to consume all of the ARC cache
pushing out data and hurting performance.  The cache is aggressively
reclaimed but this is a soft and not a hard limit.  The cache may
exceed the set limit briefly before being brought under control.

By default 25% of the ARC capacity can be used for meta-data.  This
limit can be tuned by setting the 'zfs_arc_meta_limit' module option.
Once this limit is exceeded meta-data reclaim will occur in 3 percent
chunks, or may be tuned using 'arc_reduce_dnlc_percent'.

Closes #193
2011-04-21 13:49:31 -07:00
Gunnar Beutner 36df284366 Fixed a use-after-free bug in zfs_zget().
Fixed a bug where zfs_zget could access a stale znode pointer when
the inode had already been removed from the inode cache via iput ->
iput_final -> ... -> zfs_zinactive but the corresponding SA handle
was still alive.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #180
2011-04-21 13:48:01 -07:00
Darik Horn 5b8f76ea16 Make the SPL kernel messages consistent with ZFS.
Change the SPL kernel messages for module loading and module
unloading so that they are similar to the ZFS kernel messages.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-21 09:41:13 -07:00
Darik Horn ad35b6a6e9 Remove the gawk dependency.
This reverts commit 1814251453.

Demote the gawk call back to awk and ensure that stderr is attached.  GNU gawk
tolerates a missing stderr handle, but many utilities do not, which could be
why a regular awk call was unexplainably failing on some systems.

Use argv[0] instead of sh_path for consistency internally and with other Linux
drivers.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-21 09:41:09 -07:00
Darik Horn fa6f7d8f9d Import spl_hostid as a module parameter.
Provide a call_usermodehelper() alternative by letting the hostid be passed as
a module parameter like this:

  $ modprobe spl spl_hostid=0x12345678

Internally change the spl_hostid variable to unsigned long because that is the
type that the coreutils /usr/bin/hostid returns.

Move the hostid command into GET_HOSTID_CMD for consistency with the similar
GET_KALLSYMS_ADDR_CMD invocation.

Use argv[0] instead of sh_path for consistency internally and with other Linux
drivers.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-21 09:41:01 -07:00
Brian Behlendorf 3dfc591ac4 Linux 2.6.39 compat, zlib_deflate_workspacesize()
The function zlib_deflate_workspacesize() now take 2 arguments.
This was done to avoid always having to allocate the maximum size
workspace (268K).  The caller can now specific the windowBits and
memLevel compression parameters to get a smaller workspace.

For our purposes we introduce a spl_zlib_deflate_workspacesize()
wrapper which accepts both arguments.  When the two argument
version of zlib_deflate_workspacesize() is available the arguments
are passed through.  When it's not we assume the worst case and
a maximally sized workspace is used.
2011-04-20 14:39:15 -07:00
Brian Behlendorf b1cbc4610c Linux 2.6.39 compat, kern_path_parent()
The path_lookup() function has been renamed to kern_path_parent()
and the flags argument has been removed.  The only behavior now
offered is that of LOOKUP_PARENT.  The spl already always passed
this flag so dropping the flag does not impact us.
2011-04-20 12:30:17 -07:00
Brian Behlendorf 83c623aa1a Linux 2.6.39 compat, DEFINE_SPINLOCK()
This is a long over due compatibility change.  Way, way, way back
in 2007 there was a push to remove all consumers of SPIN_LOCK_UNLOCKED.
Finally, in 2011 with 2.6.39 all the consumers have been updated
and SPIN_LOCK_UNLOCKED was removed.  It's about time we use the
new API as well, this change does exactly that.  DEFINE_SPINLOCK()
was available as far back as 2.6.12 so there doesn't need to be
any additional autoconf-foo for this change.
2011-04-20 12:01:11 -07:00
Brian Behlendorf d247f2a3cc Suppress 'zfs receive' memory warning
As part of zfs_ioc_recv() a zfs_cmd_t is allocated in the kernel
which is 17808 bytes in size.  This sort of thing in general should
be avoided.  However, since this should be an infrequent event for
now we allow it and simply suppress the warning with the KM_NODEBUG
flag.  This can be revisited latter if/when it becomes an issue.

Closes #178
2011-04-20 10:22:31 -07:00
Aniruddha Shankar 9caef54224 Added required runlevel info for init on Debian
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #208
2011-04-20 09:55:57 -07:00
Brian Behlendorf 3fce1d0962 Update zconfig.sh to use new zvol names
This change should have occured when we commited the new udev
rules for zvols.  Basically, the test script is just out of date.
We need to update it to use the /dev/zvol/ device names, and
to expect the more common -partN suffixes.

I added a udev_trigger() call in zconfig_partition() and
zconfig_zvol_device_stat() to ensure that all the udev rules have
run before.  This ensures the devices are available to subsequent
commands and closes a small race.

Finally, I was forced added a small 'sleep 1' to test 10.  I
was observing occassional failures in my VM due to the device
still claiming to be busy.  Delaying betwen the various methods
of adding/removing a vdev avoids the issue.

Closes #207
2011-04-19 16:33:41 -07:00
Brian Behlendorf 18f4238277 Add parted and lsscsi dependencies to zfs-test
The zfault.sh and zconfig.sh test scripts requires the parted
utility, the lsscsi utility, and the scsi_debug module.  To
ensure the utilities are available they have been added as
dependencies to zfs-test package.  Checking for scsi_debug
is a little more problematic because if it's missing you will
need to build it.  For clarity the documention has been updated
to mention this.

Closes #205
Closes #206
2011-04-19 15:22:46 -07:00
Brian Behlendorf cbb1e401f9 Add Gunnar Beutner to AUTHORS for his contributions 2011-04-19 14:14:51 -07:00
Gunnar Beutner bec30953cd Truncate the xattr znode when updating existing attributes.
If the attribute's new value was shorter than the old one the old
code would leave parts of the old value in the xattr znode.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #203
2011-04-19 14:14:40 -07:00
Gunnar Beutner 274b7e79f3 Added missing initialization for va.va_dentry in zfs_get_xattrdir.
Without this we may mistakenly believe we have a dentry and try to
d_instantiate() it.  This will result in the following BUG.  It's
important to note that while the xattr directory has an inode
assoicated with it we never create a dentry for it.

  kernel BUG at fs/dcache.c:1418!

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #202
2011-04-19 13:57:41 -07:00
Richard Laager 826ab7ad19 Support IEC base-2 prefixes
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 12:56:21 -07:00
Richard Laager 4da4a9e1a7 Cleanup various Sun/Solaris-isms
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 11:42:42 -07:00
Richard Laager 251eb26d17 Update the version in the zpool upgrade example
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 11:40:45 -07:00
Richard Laager 3b2041509f Normalize the deferred destruction language
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 11:40:26 -07:00
Richard Laager 0d122e21ff Improve the wording about hot spares
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 11:39:37 -07:00
Richard Laager 6b92390fe8 Improve some quoting consistency
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 11:38:56 -07:00
Richard Laager 577468215b Remove a stray tab
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 11:38:27 -07:00
Richard Laager 25d4782bac Linux has "partitions", not "slices"
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 11:38:04 -07:00
Richard Laager 54e5f2264d Use Linux disk names in zpool.8
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 11:37:31 -07:00
Richard Laager 1dc3fea59e More and correct an example in zpool.8
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 11:36:17 -07:00
Richard Laager 1fe2e23771 Change /dev/dsk -> /dev in the man pages
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 11:35:21 -07:00
Richard Laager 2d1b7b0b97 Correct man page section numbers for Linux
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2011-04-19 11:33:51 -07:00
Brian Behlendorf 12c1acde76 Set -Wno-unused-but-set-variable globally
As of gcc-4.6 the option -Wunused-but-set-variable is enabled by
default.  While this is a useful warning there are numerous places
in the ZFS code when a variable is set and then only checked in an
ASSERT().  To avoid having to update every instance of this in the
code we now set -Wno-unused-but-set-variable to suppress the warning.

Additionally, when building with --enable-debug and -Werror set these
warning also become fatal.  We can reevaluate the suppression of these
error at a later time if it becomes an issue.  For now we are basically
just reverting to the previous gcc behavior.
2011-04-19 10:44:10 -07:00
Brian Behlendorf 79713039a2 Fix gcc configure warnings
Newer versions of gcc are getting smart enough to detect the sloppy
syntax used for the autoconf tests.  It is now generating warnings
for unused/undeclared variables.  Newer version of gcc even have
the -Wunused-but-set-variable option set by default.  This isn't a
problem except when -Werror is set and they get promoted to an error.
In this case the autoconf test will return an incorrect result which
will result in a build failure latter on.

To handle this I'm tightening up many of the autoconf tests to
explicitly mark variables as unused to suppress the gcc warning.
Remember, all of the autoconf code can never actually be run we
just want to get a clean build error to detect which APIs are
available.  Never using a variable is absolutely fine for this.

Closes #176
2011-04-19 10:10:47 -07:00
Brian Behlendorf 98e2afd1c5 Fix unused variable
Flagged by the default -Wunused-but-set-variable gcc option when
running under Fedora 15.  Since it's correct this variable is
entirely unused this commit removes it.
2011-04-19 09:45:36 -07:00
Brian Behlendorf 03318641af Fix gcc configure warnings
Newer versions of gcc are getting smart enough to detect the sloppy
syntax used for the autoconf tests.  It is now generating warnings
for unused/undeclared variables.  Newer version of gcc even have
the -Wunused-but-set-variable option set by default.  This isn't a
problem except when -Werror is set and they get promoted to an error.
In this case the autoconf test will return an incorrect result which
will result in a build failure latter on.

To handle this I'm tightening up many of the autoconf tests to
explicitly mark variables as unused to suppress the gcc warning.
Remember, all of the autoconf code can never actually be run we
just want to get a clean build error to detect which APIs are
available.  Never using a variable is absolutely fine for this.
2011-04-19 09:41:41 -07:00
Brian Behlendorf 9b0f9079d2 Linux 2.6.39 compat, invalidate_inodes()
To resolve a potiential filesystem corruption issue a second
argument was added to invalidate_inodes().  This argument controls
whether dirty inodes are dropped or treated as busy when invalidating
a super block.  When only the legacy API is available the second
argument will be dropped for compatibility.
2011-04-19 09:08:08 -07:00
Brian Behlendorf 03514b0110 Fix gcc compiler warning, parse_option()
When compiling ZFS in user space gcc-4.6.0 correctly identifies
the variable 'value' as being set but never used.  This generates a
warning and a build failure when using --enable-debug.  Once again
this is correct but I'm reluctant to remove 'value' because we are
breaking the string in to name/value pairs.  While it is not used
now there's a good chance it will be soon and I'd rather not have
to reinvent this.  To suppress the warning with just as a VERIFY().
This was observed under Fedora 15.

  cmd/mount_zfs/mount_zfs.c: In function ‘parse_option’:
  cmd/mount_zfs/mount_zfs.c:112:21: error: variable ‘value’ set but not
  used [-Werror=unused-but-set-variable]
2011-04-19 09:04:51 -07:00
Brian Behlendorf 0fe3d820f5 Fix gcc compiler warning, dsl_pool_create()
When compiling ZFS in user space gcc-4.6.0 correctly identifies
the variable 'os' as being set but never used.  This generates a
warning and a build failure when using --enable-debug.  However,
the code is correct we only want to use 'os' for the kernel space
builds.  To suppress the warning the call was wrapped with a
VERIFY() which has the nice side effect of ensuring the 'os'
actually never is NULL.  This was observed under Fedora 15.

  module/zfs/dsl_pool.c: In function ‘dsl_pool_create’:
  module/zfs/dsl_pool.c:229:12: error: variable ‘os’ set but not used
  [-Werror=unused-but-set-variable]
2011-04-19 09:04:51 -07:00
Brian Behlendorf e30c0ada6d Linux 2.6.39 compat, invalidate_inodes()
Update code to use the spl_invalidate_inodes() wrapper.  This hides
some of the complexity of determining if invalidate_inodes() was
exported, and if so what is its prototype.  The second argument
of spl_invalidate_inodes() determined the behavior of how dirty
inodes are handled.  By passing a zero we are indicated that we
want those inodes to be treated as busy and skipped.
2011-04-19 08:57:23 -07:00
Brian Behlendorf cbf81d4c3b Autogen refresh for kernel-insert-inode-locked.m4
Several Makefile.in's were accidentally not updated when the
kernel-insert-inode-locked.m4 check was added.  This change simply
refreshes the missed files.
2011-04-18 15:47:18 -07:00
Brian Behlendorf 0f6fd32927 Fix rebuildable RPMs for el6/ch5
When rebuilding the source RPM under el5 you need to append the
target_cpu.  However, under el6/ch5 things are packaged correctly
and the arch is already part of kver.  For this reason it also
needs to be stripped from kver when setting kverpkg.
2011-04-08 10:22:42 -07:00
Brian Behlendorf 96cdefab84 Fix rebuildable RPMs for el6/ch5
When rebuilding the source RPM under el5 you need to append the
target_cpu.  However, under el6/ch5 things are packaged correctly
and the arch is already part of kver.  For this reason it also
needs to be stripped from kver when setting kverpkg.
2011-04-08 10:20:08 -07:00
Ned Bass 38baef530d Align closing fi in mount-zfs.sh 2011-04-08 10:05:18 -07:00