ngie [Wed, 13 May 2015 10:53:16 +0000 (10:53 +0000)]
MFC r281618,r281823:
r281618:
Defeat race with MK_KERBEROS == yes introduced with bootstrap-tools
parallelization work done in r279197
- kerberos5/lib/libroken requires kerberos5/tools/make-roken to build
- kerberos5/tools/asn1_compile, kerberos5/tools/slc, and usr.bin/compile_et
require kerberos5/lib/libroken and kerberos5/lib/libvers
This race is incredibly evident when cross-building sparc64 on
ref10-amd64.freebsd.org
Pointyhat to: ngie
r281823:
Serialize all of _kerberos5_bootstrap_tools to avoid build failures involving
make bootstrap-tools
On the plus side, this also greatly reduces complexity
Pointyhat to: ngie
Reported by: Willem Jan Withagen <wjw@digiware.nl>
mav [Wed, 13 May 2015 10:35:34 +0000 (10:35 +0000)]
MFC r282364, r282429, r282524:
Implement BSY reporting and in-order execution of non-NCQ commands.
Block command queue on BSY or DRQ bits set or device error reported.
This can be a performance penalization for non-NCQ commands, but it is
required for proper error recovery and standard compliance.
ngie [Wed, 13 May 2015 10:29:02 +0000 (10:29 +0000)]
MFC r282073:
- Fix style(9) a bit
-- Fix whitespace
-- Use err/errx
-- Remove superfluous braces
- Be a bit more defensive with input from the end-user
- Don't throw a floating point exception by dividing by 0 when processing a
zero-byte file
hiren [Wed, 13 May 2015 08:04:50 +0000 (08:04 +0000)]
MFC r281984:
Currently there is no easy way to specify net.isr.maxthreads = all cpus. We need
to specify exact number of cpus in loader.conf which get annoying when you have
mix of machines which don't have equal number of total cpus. I propose "-1" as
that value. When loader.conf has net.isr.maxthreads = -1, netisr will use all
available cpus.
hiren [Wed, 13 May 2015 07:39:16 +0000 (07:39 +0000)]
Partial MFC r281838:
For igb(4), when we are doing multiqueue, we are all setup to have full 32bit
RSS hash from the card. Expose that so others like lagg(4) can use that and
avoid hashing the traffic by themselves. Setting hashtype as OPAQUE because
FreeBSD 10 doesn't have RSS support.
emaste [Tue, 12 May 2015 18:09:54 +0000 (18:09 +0000)]
MFC r281944: vt(4): Simplify mouse area detection
vt_is_cursor_in_area needs to return true if any part of the mouse
cursor is visible in the rectangle area. Replace the existing test with
a simpler version of a test for overlapping rectangles.
Previously the mask wrapped when one or more of the mask bytes extended
past the right edge of the window. Simplify the logic and use the same
byte offset and bit in both the pattern and mask.
gnn [Sat, 9 May 2015 19:36:30 +0000 (19:36 +0000)]
MFC: 281529
I can find no reason to allow packets with both SYN and FIN bits
set past this point in the code. The packet should be dropped and
not massaged as it is here.
mav [Wed, 6 May 2015 21:08:16 +0000 (21:08 +0000)]
MFC r281860: Make AIO to not allocate pbufs for unmapped I/O like r281825.
While there, make few more performance optimizations.
On 40-core system doing many 512-byte AIO reads from array of raw SSDs
this change removes lock congestions inside pbuf allocator and devfs,
and bottleneck on single AIO completion taskqueue thread. It improves
peak AIO performance from ~600K to ~1.3M IOPS.
mav [Wed, 6 May 2015 21:06:32 +0000 (21:06 +0000)]
MFC r281825: Rewrite physio() to not allocate pbufs for unmapped I/O.
pbufs is a limited resource, and their allocator is not SMP-scalable.
So instead of always allocating pbuf to immediately convert it to bio,
allocate bio just here. If buffer needs kernel mapping, then pbuf is
still allocated, but used only as a source of KVA and storage for a list
of held pages.
On 40-core system doing many 512-byte reads from user level to array of
raw SSDs this change removes huge lock congestion inside pbuf allocator.
It improves peak performance from ~300K to ~1.2M IOPS. On my previous
24-core system this problem also existed, but was less serious.
Update the Linux compatibility layer:
- Add more functions.
- Add some missing includes which are needed when the header files
are not included in a particular order.
- The kasprintf() function cannot be inlined due to using a variable
number of arguments. Move it to a C-file.
- Fix problems about 32-bit ticks wraparound and unsigned long
conversion. Jiffies or ticks in FreeBSD have integer type and are
not long.
- Add missing "order_base_2()" macro.
- Fix BUILD_BUG_ON() macro.
- Declare a missing symbol which is needed when compiling without -O2
- Clean up header file inclusions in the linux/completion.h, linux/in.h
and linux/fs.h header files.
hselasky [Tue, 5 May 2015 19:56:23 +0000 (19:56 +0000)]
MFC r281881:
Disable multi process interrupts, because the current code doesn't use
them. Else we can end up in an infinite interrupt loop in USB device
mode.
hselasky [Tue, 5 May 2015 19:47:17 +0000 (19:47 +0000)]
MFC r282120:
The add_bounce_page() function can be called when loading physical
pages which pass a NULL virtual address. If the BUS_DMA_KEEP_PG_OFFSET
flag is set, use the physical address to compute the page offset
instead. The physical address should always be valid when adding
bounce pages and should contain the same page offset like the virtual
address.
markj [Tue, 5 May 2015 03:13:02 +0000 (03:13 +0000)]
MFC 281701:
SDT(9): add a section on SDT providers, mentioning the "sdt" provider.
Add examples demonstrating how one can list available providers and the
DTrace probes provided by a provider.
mav [Mon, 4 May 2015 19:33:51 +0000 (19:33 +0000)]
MFC r281765:
Activate write-only optimization if bpf device opened with O_WRONLY.
dhclient opens bpf as write-only to send packets. It never reads received
packets from that descriptor, but processing them in kernel takes time.
Especially much time takes packet timestamping on systems with expensive
timecounter, such as bhyve guest, where network speed dropped in half.
kib [Mon, 4 May 2015 08:16:32 +0000 (08:16 +0000)]
MFC r282085:
Partially revert r255986: do not call VOP_FSYNC() when helping
bufdaemon in getnewbuf(), do use buf_flush(). The difference is that
bufdaemon uses TRYLOCK to get buffer locks, which allows calls to
getnewbuf() while another buffer is locked.
Make sure the adapter's management queue and the event queue are
available before any uppper layer driver (TOE, iWARP, or iSCSI)
registers with the base cxgbe(4) driver.
cxgbe(4): Verify that the addresses in if_multiaddrs really are multicast
addresses. (The chip doesn't really care, it's just that it needs to be
told explicitly if unicast DMACs are checked for "hits" in the hash that
is used after the TCAM entries are all used up).
mav [Sun, 3 May 2015 07:13:14 +0000 (07:13 +0000)]
MFC r281026, r281108, r281109:
Make ZFS ARC track both KVA usage and fragmentation.
Even on Illumos, with its much larger KVA, ZFS ARC steps back if KVA usage
reaches certain threshold (3/4 on i386 or 16/17 otherwise). FreeBSD has
even less KVA, but had no such limit on archs with direct map as amd64.
As result, on machines with a lot of RAM, during load with very small user-
space memory pressure, such as `zfs send`, it was possible to reach state,
when there is enough both physical RAM and KVA (I've seen up to 25-30%),
but no continuous KVA range to allocate even single 128KB I/O request.
Address this situation from two sides:
- restore KVA usage limitations in a way the most close to Illumos;
- introduce new requirement for KVA fragmentation, specifying that we
should have at least one sequential KVA range of zfs_max_recordsize bytes.
Experiments show that first limitation done alone is not sufficient. On
machine with 64GB of RAM it is sometimes needed to drop up to half of ARC
size to get at leats one 1MB KVA chunk. Statically limiting ARC to half
of KVA/RAM is too strict, so second limitation makes it to work in cycles:
accumulate trash up to certain critical mass, do massive spring-cleaning,
and then start littering again.
rmacklem [Sat, 2 May 2015 12:18:28 +0000 (12:18 +0000)]
MFC: r281962
Fix the NFS server's handling of a bogus NFSv2 ROOT RPC.
The ROOT RPC is deprecated in the NFSv2 RFC, RFC-1094
and should never be used by a client.
mav [Fri, 1 May 2015 17:30:59 +0000 (17:30 +0000)]
MFC r281700:
Workaround bhyve virtual disks operation on top of GEOM providers.
GEOM does not support scatter/gather lists in its I/Os. Such requests
are cut in pieces by physio(), that may be problematic, if those pieces
are not multiple of provider's sector size. If such case is detected,
move the data through temporary sequential buffer.
MFC: r281628
mav@ has found that NFS servers exporting ZFS file systems
can perform better when using a 128K read/write data size.
This patch changes NFS_MAXDATA from 64K to 128K so that
clients can use 128K for NFS mounts to allow this.
The patch also renames NFS_MAXDATA to NFS_SRVMAXIO so
that it is clear that it applies to the NFS server side
only. It also avoids a name conflict with the NFS_MAXDATA
defined in rpcsvc/nfs_prot.h, that is used for userland RPC.
MFC: r281562
File systems that do not use the buffer cache (such as ZFS) must
use VOP_FSYNC() to perform the NFS server's Commit operation.
This patch adds a mnt_kern_flag called MNTK_USES_BCACHE which
is set by file systems that use the buffer cache. If this flag
is not set, the NFS server always does a VOP_FSYNC().
This should be ok for old file system modules that do not set
MNTK_USES_BCACHE, since calling VOP_FSYNC() is correct, although
it might not be optimal for file systems that use the buffer cache.
MFC 280721:
Allow additional flags to be passed to netstat -i in the daily status check.
In particular, this allows an administrator to specify "-h" for human
readable output if that is preferred.
The default setting passes "-d", so that can be excluded by using a custom
setting.
drm: Update the device-independent code to match Linux 3.8.13
This update brings few features:
o Support for the setmaster/dropmaster ioctls. For instance, they
are used to run multiple X servers simultaneously.
o Support for minor devices. The only user-visible change is a new
entry in /dev/dri but it is useless at the moment. This is a
first step to support render nodes [1].
The main benefit is to greatly reduce the diff with Linux (at the
expense of an unreadable commit diff). Hopefully, next upgrades will be
easier.
No updates were made to the drivers, beside adapting them to API
changes.
Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
Date: Fri May 31 12:17:08 2013 +0000
drm: Sort connector modes based on vrefresh
Keeping the modes sorted by vrefresh before the pixel clock makes the
mode list somehow more pleasing to the eye.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
PR: 198936 (r280814)
Tested by: Many people
MFC of: r280183, r280187 (original commit by glebius), r280814
Relnotes: yes