This was the cause of poor ZFS performance in misc-fs-stress test.

Before:

 Wrote 168.129 MB in 10.12 s = 16.610 Mb/s
 Wrote 194.688 MB in 10.00 s = 19.469 Mb/s
 Wrote 183.004 MB in 10.06 s = 18.186 Mb/s
 Wrote 167.754 MB in 10.28 s = 16.315 Mb/s

After:

 Wrote 636.227 MB in 10.00 s = 63.623 Mb/s
 Wrote 666.979 MB in 10.00 s = 66.696 Mb/s
 Wrote 613.512 MB in 10.00 s = 61.350 Mb/s
 Wrote 573.502 MB in 10.00 s = 57.346 Mb/s
 Wrote 668.607 MB in 10.00 s = 66.857 Mb/s
 Wrote 630.920 MB in 10.00 s = 63.087 Mb/s

It turned out that the limiting factor was the ARC cache. A check
inside arc_tempreserve_space() was forcing txg to be synced too often
(once every 400ms). The arc_c variable was only 16M (arc_c_min) which
allowed to write only 8M per transaction. It turns out that arc_c
depends on kmem_size() which is based on physmem which was never
initialized.

I would hold with commiting this yet because of several reasons,
which I want to put under your consideration.

While this improves write throughput it makes the boot time after make
much longer, on my disk the boot time is increased from 1.5s to 10s.
This is because zfs verifies the last 3 txgs upon mount. This patch
increases txg size, which results in more data to check in the next
boot. I'm working on solving this right now.

Something worth noting is that while larger transactions sync less
often incresing throughput they also sync longer increasing worst case
latency. In my test the pauses get as high as 3 seconds with 1G of
guest memory.

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

While (unfortunately) C++ doesn't support designated initializers, and the
compiler rejects them, one instance has survived in xenbus.  Strangely,
gcc 4.8.2 generates correct code, while gcc 4.8.0 fails with an internal
compiler error, instead of both of them rejecting the code.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Backported from FreeBSD r242252.

Improves netperf by about 10%.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

This is more useful if there is no ordering between the two numbers (either
one can be ahead).

Change BYTES_THIS_ACK to return unsigned, to prevent an unsigned division from
turning into a signed division.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Add comparison operators that use modulo arithmetic to order sequence
numbers, and use them to replace SEQ_LT() and friends, increasing code
readability.

As a consequence std::min() and std::max() can be used instead of SEQ_MIN()
and SEQ_MAX().

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

tcp sequence numbers are similar to integers, but have different comparison
operations.  Separate them into a class so we don't mix the two.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

inline functions can be overloaded and are less nasty than macros in other
ways (like evaluating their arguments only once).

Note we can't touch ntohl() itself, since it is defined to be an out-of-line
function by libc.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Helps making bsd header changes that xen includes.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

We use sched::thread::attr to pass parameters to sched::thread creation,
i.e., create a thread with non-default stack parameters, pinned to a
particular CPU, or a detached thread.

Previously we had constructors taking many combinations of stack size
(integer), pinned cpu (cpu*) and detached (boolean), and doing "the
right thing". However, this makes the code hard to read (what does
attr(4096) specify?) and the constructors hard to expand with new
parameters.

Replace the attr() constructors with the so-called "named parameter"
idiom: attr now only has a null constructor attr(), and one modifies
it with calls to pin(cpu*), detach(), or stack(size).

For example,
    attr()                                  // default attributes
    attr().pin(sched::cpus[0])              // pin to cpu 0
    attr().stack(4096).pin(sched::cpus[0])  // pin and non-default stack
    and so on.

Signed-off-by: Nadav Har'El <nyh@cloudius-systems.com>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

This patch applies bugfix published on FreeBSD list at Feb 2013:
http://lists.freebsd.org/pipermail/svn-src-stable-9/2013-February/003928.html



LRO mechanism is broken on systems without IP checksum verification offload.
Due to improper checksum verification RX packets omit LRO path and go
directly to TCP stack which is not good for performance.

EC2 Xen is one example of such a system.
This bug is one of the reasons we see bad performance on Amazon.

Some test results w/ and w/o the fix:

Buffer size    Before         After          Improvement %
TCP TX
32             557.52         1386.28        149
64             552.38         1385.99        151
128            546.43         1401.46        156
256            565.25         1382.28        145
512            557.32         1375.23        147
1024           549.71         1356.69        147
2048           551.11         1371.92        149
4096           556.13         1383.67        149
8192           559.49         1364.05        144
16384          567.25         1366.48        141
32768          546.18         1366.63        150
65536          553.4          1353.87        145

TCP RX
32             107.37         105.48         -2
64             187.56         179.9          -4
128            297.16         301.71         2
256            300.47         503.92         68
512            294.76         826.13         180
1024           299.95         1916.69        539
2048           287.04         1924.44        570
4096           300.78         1929.37        541
8192           304.52         1934.02        535
16384          305.04         1957.54        542
32768          309            1921.84        522
65536          296.48         1935.41        553

Still we are pretty far from Linux, there are other problems to be fixed.

Signed-off-by: Dmitry Fleytman <dmitry@daynix.com>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

if_getinfo is shorter and more consistent all the other if_ functions
name, e.g., if_init, if_ioctl.

Signed-off-by: Asias He <asias@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Error message:

bsd/sys/kern/uipc_socket.cc: In function 'socket* soalloc(vnet*)':
bsd/sys/kern/uipc_socket.cc:245:58: error: expected ')' before 'PRIx64'
  uipc_d("soalloc() so=%" PRIx64, (uint64_t)so);
                                                          ^

This was introduced in one of these commits:
c506516b14d32e416f4a9247869a1a89f7e9c413
406a93ed

As suggested by Pekka including <cinttypes> is the proper fix for C++11.

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

It's just to painful to do this piecemeal.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Added ifnet->if_get_if_info() method to collect the
internally gathered statistics into the standard if_data struct.

Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Reviewed-by: Asias He <asias@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Added IFCAP_HWSTATS iface capability indicating that
device driver and HW are capable to handle the statistics internally.

Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Reviewed-by: Asias He <asias@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Over the process of creating a zpool, history_str_get() will be eventually
called by zfs_ioc_pool_create(). history_str_get() will not work as expected as
it fully relies on the function copyinstr() which is currently unimplemented.

After that, spa_create() will be called with history_str storing trash, thus
making the spa_history_log filled with completely wrong entries.

Let's fix this problem simply by implementing copyinstr. By the way, copystr
has basically the same semantics as copyinstr, so let's implement it as well.
It's not being used anywhere, but who knows? It might be needed in the future
by some bsd compliant application.

As far as I know, this problem didn't manifest *visibly* anywhere as the spa
history log wasn't intentionally used by us yet. I found the problem while
auditing code.

Signed-off-by: Raphael S. Carvalho <raphaelsc@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

bsd defines some m_ macros, for example m_flags, to save some typing.  However
if you have a variable of the same name in another header, for example
m_flags, have fun trying to compile your code.

Expand the code in place and eliminate the macros.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

The bsd specific ones are nenumbered to avoid clashes.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Since most SOL_* macros are equivalent to the IPPROTO_* defines, only one
define needs to be changed.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Move them to a common include file.

Since they're defined externally, there is no real conflict.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Unused.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Fix a unicode character in the author's name.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Fix some compilation errors that would arrise when NDEBUG is defined:
 - tests/misc-tcp.cc: missing #include<iostream> that would come from
   include/boost/assert.hpp: line 81 when NDEBUG is not defined.
 - bsd/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_dataset.c:
   when NDEBUG is not defined assert() is defined to __assert_fail,
   which has "noreturn" attribute, which satisfies the compiler.
   When NDEBUG is defined and assert() completly compiles out, the
   compiler start complaining about the missing "return" statment.

Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Currently, namei() does vget() unconditionally if no dentry is found.
This is wrong because the path can be a hard link that points to a vnode
that's already in memory.

To fix the problem:

  - Use inode number as part of the hash in vget()

  - Use vn_lookup() in vget() to make sure we have one vnode in memory
    per inode number.

  - Push the vget() calls down to individual filesystems and make
    VOP_LOOKUP return an vnode

Changes since v2:
  - v1 dropped lock in vn_lookup, thus assert that vnode_lock is held.

Changes since v3:
  - Fix lock ordering issue in dentry_lookup. The lock respective to the parent
node must be acquired before dentry_lookup and released after the process is
done. Otherwise, a second thread looking up for the same dentry may take the
'NULL' path incorrectly.

Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>
Signed-off-by: Raphael S. Carvalho <raphaelsc@cloudius-systems.com>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

This patch fixes the error codes in four error cases:

1. unlink() of a directory used to return EPERM (as in Posix), and now
   returns EISDIR (as in Linux).

2. rmdir() of a non-empty directory used to return EEXIST (as in Posix)
   and now returns ENOTEMPTY (as in Linux).

3. rmdir() of a regular file (non-directory) used to return EBADF
   and now returns ENOTDIR (as in Linux).

4. readdir() of a regular file (non-directory) used to return EBADF
   and now returns ENOTDIR (as in Linux).

This patch also adds a test, tst-remove.cc, for the various unlink() and
rmdir() success and failure modes.

Fixes #123.

Signed-off-by: Nadav Har'El <nyh@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

This reverts commit e4aad1ba.

It causes tst-vfs.so to just hang.

Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Currently, namei() does vget() unconditionally if no dentry is found.
This is wrong because the path can be a hard link that points to a vnode
that's already in memory.

To fix the problem:

  - Use inode number as part of the hash in vget()

  - Use vn_lookup() in vget() to make sure we have one vnode in memory
    per inode number.

  - Push the vget() calls down to individual filesystems and make
    VOP_LOOKUP return an vnode

  - Drop lock in vn_lookup() and assert that vnode_lock is held.

Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>
Signed-off-by: Raphael S. Carvalho <raphaelsc@cloudius-systems.com>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

OSv does not support xattrs in ZFS. Fix up a build breakage reported by Glauber:

  /home/ubuntu/osv/bsd/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c: In function ‘zfs_setattr’:
  /home/ubuntu/osv/bsd/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c:2152:12: error: ‘xoap’ may be used uninitialized in this function [-Werror=uninitialized]

Reported-by: Glauber Costa <glommer@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Probably the worst indented file in the whole tree.  Set Eclipse Ctrl-I
on it.

Also fix a unicode character in the author's name.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Prior to 65ccda4c (net: use a file derived class for sockets
 (socket_file)), ioctl()s for socket were directed to linux_ioctl_socket()
and thence to soo_ioctl().  However that commit short-circuited
linux_ioctl_socket() out and dipatched directly to what was previously
known as soo_ioctl() (and became socket_file::ioctl()).  The caused
interface enumeration ioctl()s to fail, for example in Cassandra.

Fix by bringing back the previous behaviour.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

The only caller, soo_close() can only be called from a context where
no file references remain, so no further file API calls can be made.

Remove it.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>