David Malone [Tue, 23 Feb 2016 15:28:13 +0000 (15:28 +0000)]
Following revision r295924, the changes to a db file should be fsynced
before the file is closed. Consequently, it shouldn't be necessary to
open the file with O_SYNC any more.
This improves the performance of building large .db files for large
password files a lot and should resolve this problem:
David Malone [Tue, 23 Feb 2016 15:21:13 +0000 (15:21 +0000)]
If we close or sync a hash-based db file, make sure to call fsync to
make sure the changes are on disk. The people at pfSense noticed that
it didn't always make it to the disk soon enough with soft updates.
Sepherosa Ziehau [Tue, 23 Feb 2016 09:25:20 +0000 (09:25 +0000)]
hyperv/hn: Use IFQ_DRV_PREPEND instead of IF_PREPEND
IF_PREPEND promises out-of-order packet sending when the TX desc list
is depleted. It was overlooked and copied blindly when the transmission
path was partially rewritten.
Ed Schouten [Tue, 23 Feb 2016 09:22:00 +0000 (09:22 +0000)]
Make handling of mmap()'s prot argument more strict.
- Make the system call fail if prot contains bits other than read, write
and exec.
- Similar to OpenBSD's W^X, don't allow write and exec to be set at the
same time. I'd like to see for now what happens if we enforce this
policy unconditionally. If it turns out that this is far too strict,
we'll loosen this requirement.
Xin LI [Tue, 23 Feb 2016 07:04:54 +0000 (07:04 +0000)]
Apply upstream commit 6e06b1c8 (partial, by kientzle):
Fix a potential crash issue discovered by Alexander Cherepanov:
It seems bsdtar automatically handles stacked compression. This is a
nice feature but it could be problematic when it's completely
unlimited. Most clearly it's illustrated with quines:
Ian Lepore [Tue, 23 Feb 2016 02:57:45 +0000 (02:57 +0000)]
Now that we have OF_decode_addr(), with proper MD implementations, to obtain
the bus space tag and handle for a uart console, use the values returned by
that routine to set the global uart_bus_space_mem, instead of assuming that
there will be a global variable named fdtbus_bs_tag to set it from.
Also, use OF_getencprop() instead calling fdt32_to_cpu() separately.
Marius Strobl [Tue, 23 Feb 2016 01:19:26 +0000 (01:19 +0000)]
Fix and clean up usage of DMA and TSO segments:
- At Intel it is believed that most of their products support "only"
40 DMA segments so lower {EM,IGB}_MAX_SCATTER accordingly. Actually,
40 is more than plenty to handle full size TSO packets so it doesn't
make sense to further distinguish between MAC variants that really
can do 64 DMA segments. Moreover, capping at 40 DMA segments limits
the stack usage of {em,igb}_xmit() that - given the rare use of more
than these - previously hardly was justifiable, while still being
sufficient to avoid the problems seen with em(4) and EM_MAX_SCATTER
set to 32.
- In igb(4), pass the actually supported TSO parameters up the stack.
Previously, the defaults set in if_attach_internal() were applied,
i. e. a maximum of 35 TSO segments, which made supporting more than
these in the driver pointless. However, this might explain why no
problems were seen with IGB_MAX_SCATTER at 64.
- In em(4), take the 5 m_pullup(9) invocations performed by em_xmit()
in the TSO case into account when reporting TSO parameters upwards.
In the worst case, each of these calls will add another mbuf and,
thus, the requirement for an additional DMA segment. So for best
performance, it doesn't make sense to advertize a maximum of TSO
segments that typically will require defragmentation in em_xmit().
Again, this leaves enough room to handle full size TSO packets.
- Drop TSO macros from if_lem.h given that corresponding MACS don't
support TSO in the first place.
Ian Lepore [Tue, 23 Feb 2016 01:08:39 +0000 (01:08 +0000)]
Make imx6 systems work again after recent import of new dts files.
Linux-driven changes to the way the chip's two interrupt controllers are
defined (we only support one of them) led to no interrupt processing, so
the system would hang after device instantiation. This workaround just
rewrites the FDT data on the fly to get interrupt handling back under the
control of the main GIC device.
If/when we ever support deep sleep modes that involve powering down the
main GIC, we'll have to undo this change, write a driver for the GPC-PIC,
and somehow manage the handoff of responsibilities between the two drivers
as the chip transitions in/out of deep sleep mode.
Dimitry Andric [Mon, 22 Feb 2016 22:16:32 +0000 (22:16 +0000)]
Fix a problem in ld, causing it to sometimes print messages similar to
"invalid string offset 65521 >= 27261 for section `.strtab'". for object
files produced by recent versions of clang.
In BFD's elf_create_symbuf() function, the size of the symbol buffer
('ssymbuf') is not calculated correctly, and the initial value for the
'ssym' variable is off by one, since 'ssymbuf' has shndx_count + 1
members.
Ian Lepore [Mon, 22 Feb 2016 18:53:55 +0000 (18:53 +0000)]
If the user has set a u-boot env var named rootpath, automatically
import it into the loader(8) env as dhcp.root-path, so that it overrides
any dhcp/bootp server-provided path.
Now if you have a dhcp server available you can easily net-boot a u-boot
system even if you don't control the dhcp server config, by setting just
two variables in the u-boot env:
loaderdev=net
rootpath=<nfsserverip>:<pathname>
Previously you had to either accept all the dhcp parameters from the
server without the ability to locally provide the rootpath, or you had
to forego dhcp and set more vars (ipaddr, netmask, serverip, rootpath).
urtwn: add an option to compile the driver without firmware specific code
- Add URTWN_WITHOUT_UCODE option (will disable any firmware specific code
when set).
- Do not exclude the driver from build when MK_SOURCELESS_UCODE is set
(URTWN_WITHOUT_UCODE will be enforced unconditionally).
- Do not abort initialization when firmware cannot be loaded;
behave like the URTWN_WITHOUT_UCODE option was set.
- Drop some unused variables from urtwn_softc structure.
Tested with RTL8188EU and RTL8188CUS in HOSTAP and STA modes.
Reviewed by: kevlo
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D4849
Ian Lepore [Sun, 21 Feb 2016 14:59:24 +0000 (14:59 +0000)]
Unconditionally set e_ident[OSABI]=ELFOSABI_FREEBSD in arm binary headers.
When the armv6 support was imported from a project branch, this complex
conditional logic and related #define'd values came along, but it's really
not clear what the intent of it all was. The effect, however, was that
OSABI was always set to zero, which is "UNIX System V ABI". Having the wrong
value there causes pkg(8) to avoid looking inside arm elf binaries to
determine shared-lib required/provides info for packaging.
Dimitry Andric [Sun, 21 Feb 2016 13:03:58 +0000 (13:03 +0000)]
Fix "invalid type '(null)'" usage messages in zfs(8) and zpool(8).
Currently, zfs(8) and zpool(8) print "invalid type '(null)'" or similar
messages, if you pass in invalid types, sources or column names for "zfs
get", "zfs list" and "zpool get". This is because the commands use
getsubopt(3), and in case of failure, they print 'value', which is NULL
when sub options don't match.
They should print 'suboptarg' instead, which is the documented way to
get at the non-matching sub option value.
Reviewed by: smh
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D5365
Some BIOSes ACPI bytecode needs to take (sleepable) acpi mutex for
acpi_GetInteger() execution. Intel DMAR interrupt remapping code
needs to know UID of the HPET to properly route the FSB interrupts
from the HPET, even when interrupt remapping is disabled, and the code
is executed under some non-sleepable mutexes.
Cache HPET UIDs in the device softc at the attach time and provide
lock-less method to get UID, use the method from the dmar hpet
handling code instead of calling GetInteger().
Reported and tested by: Larry Rosenman <ler@lerctr.org>
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Kristof Provost [Sat, 20 Feb 2016 12:53:53 +0000 (12:53 +0000)]
in pf_print_state_parts, do not use skw->proto to print the protocol but our
local copy proto that we very carefully set beforehands. skw being NULL is
perfectly valid there.
Zbigniew Bodek [Sat, 20 Feb 2016 12:28:20 +0000 (12:28 +0000)]
Revert r295756:
Extract common code from PowerPC's ofw_pci
Import portions of the PowerPC OF PCI implementation into
new file "ofw_pci.c", common for other platforms. The files ofw_pci.c and
ofw_pci.h from sys/powerpc/ofw no longer exist. All required declarations
are moved to sys/dev/ofw/ofw_pci.h.
This creates a new ofw_pci_write_ivar() function and modifies
ofw_pci_nranges(), ofw_pci_read_ivar(), ofw_pci_route_interrupt()
methods.
Most functions contain existing ppc implementations in the majority
unchanged. Now there is no need to have multiple identical copies
of methods for various architectures.
Kristof Provost [Sat, 20 Feb 2016 11:36:35 +0000 (11:36 +0000)]
ifconfig(8): can't use 'name' or 'description' when creating interface with auto numbering
If one does 'ifconfig tap create name blah', it will return error because the
'name' command doesn't properly populate the request sent to ioctl(...). The
'description' command has the same bug, and is also fixed with this patch.
If one does 'ifconfig tap create mtu 9000 name blah', it DOES work, but 'tap0'
(or other sequence number) is echoed, instead of the expected 'blah'. (assuming
the name change actually succeeded)
Submitted by: Marie Helene Kvello-Aune <marieheleneka@gmail.com>
Differential Revision: https://reviews.freebsd.org/D5341
Svatopluk Kraus [Sat, 20 Feb 2016 07:45:21 +0000 (07:45 +0000)]
Rename busdma_machdep.c to busdma_machdep-v4.c, pmap.c to pmap-v4.c
and trap.c to trap-v4.c to be plain and consistent with other armv4
specific files.
Eric Joyner [Fri, 19 Feb 2016 22:48:20 +0000 (22:48 +0000)]
ixl(4)/ixlv(4): Revert m_collapse() in ixl_xmit() to m_defrag().
The m_collapse() call would fail when transmitting medium-sized packets
when the interface mtu was set to 9000, so revert back to m_defrag(),
which does not fail.
Eric Joyner [Fri, 19 Feb 2016 22:45:09 +0000 (22:45 +0000)]
ixl(4): Fix errors in queue interrupt setup in MSIX mode.
- I40E_PFINT_DYN_CTLN needs to be cleared, and not have a queue index
written to it.
- The interrupt linked list for each queue is changed to only include
the queue's Rx and Tx queues.
Modified the use of bxe_grc_dump() function so that it can be invoked directly at any potential error path, where a fwdump is needed. The fwdump (a.k.a grcdump) is stored in a driver buffer. The sysctl grcdump_done indicates if a fwdump was taken and waiting to be retrieved.
The sysctl trigger_grcdump can be used to manually trigger a fwdump.
Stefan Eßer [Fri, 19 Feb 2016 16:53:21 +0000 (16:53 +0000)]
Fix syntax error introduced in previous commit where I removed one
character to few. I should have waited for the kernel compile to finish,
even though the change seemed so trivial.
Andrew Turner [Fri, 19 Feb 2016 14:15:31 +0000 (14:15 +0000)]
Add initial support for the Allwinner A31i and A31s. This just adds the
FDT platform code to detect when we are booting on one of these SoCs. The
driver changes will be added shortly.
Stefan Eßer [Fri, 19 Feb 2016 14:01:35 +0000 (14:01 +0000)]
Fix possible out-of-bounds access detected by Ulrich Spörleins "scan-build".
Some invalid PCI device selectors could cause read access to an initialized
variable next to the array (local loop index variable).
While here, the parser has been made more strict with regard to the syntax
of PCI device selectors as documented in the man-page. E.g. "pci:" used to
be interpreted as "pci0:0".
Svatopluk Kraus [Fri, 19 Feb 2016 09:23:32 +0000 (09:23 +0000)]
Rename pte.h to pte-v4.h and start including directly either pte-v4.h
or pte-v6.h in files which needs it.
There are quite internal definitions in pte-v4.h and pte-v6.h headers
specific for corresponding pmap implementation. These headers should be
included only in very few files and an intention is to not hide for
which implementation such files are.
Further, sys/arm/arm/elf_trampoline.c is an example of file which
uses armv4 like pmap implementation for both armv4 and armv6 platforms.
This is another reason why pte.h which includes specific header
according to __ARM_ARCH is not created.
Stefan Eßer [Fri, 19 Feb 2016 08:42:13 +0000 (08:42 +0000)]
Remove O_SYNC from the options passed to dbmopen().
The output file is created as a temporary file that is moved over the
existing file after completion. Thus there is no need to immediately
flush all created db records to the temporary file.
This speeds up creation of the termcap db by a factor of 40 on my
ZFS based /etc filesytem (from 25 seconds to 0.6 seconds).
I have compared multiple output files created with and without O_SYNC
and they came out identical each time. Nonetheless it might be best
to MFC this change and the similar one for services_mkdb (r295465) at
the same time when the changes to hash.c in review D5186 are merged.
Svatopluk Kraus [Fri, 19 Feb 2016 08:41:47 +0000 (08:41 +0000)]
Move common definitions from both pmap-v4.h and pmap-v6.h into pmap.h.
(1) MI interface needed for vm subsystem.
(2) MD interface created for ARM architecture to be used in files
shared by armv4 and armv6 platforms.
- Add definitions for Timing Advertisement and Control Wrapper frames.
- Refresh ieee80211_mgt_subtype_name and ieee80211_ctl_subtype_name
arrays.
- Count Timing Advertisement frames as discarded management frames in all
modes.
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D5331
Sepherosa Ziehau [Fri, 19 Feb 2016 05:03:17 +0000 (05:03 +0000)]
hyperv/hn: Add option to bind TX taskqueues to the specified CPU
It will be used to help tracking host side transmission ring selection
issue; and it will be turned on by default, once we have concrete result.
Reviewed by: adrian, Jun Su <junsu microsoft com>
Approved by: adrian (mento)
MFC after: 1 week
Sponsored by: Microsoft OSTC
Differential Revision: https://reviews.freebsd.org/D5316