The 'hostbridge' device exists to allow guests to infer msi/msix
capablity by advertising pcie capability.
Since the 'hostbridge' device isn't a true pci-to-pci bridge, and
doesn't actaully use the bridge configuration space layout, change
the header-type from type 1 to type 0 to avoid confusion.
Raise the SDHCI timeout to 10 seconds and add a sysctl to allow changing
this value at runtime.
The SD card specification says that a block write or a block erase can take
up to 250ms to complete and thus, under some circumstances, the existent 2
seconds timeout was triggering with normal usage.
This change fixes the sporadic controller timeout that happens on RPi and
RPi 2.
Jung-uk Kim [Thu, 21 May 2015 19:31:10 +0000 (19:31 +0000)]
Do not probe Intel PIIX4 south bridge quirks on amd64. These quirky south
bridges only supported Intel Pentium and Pentium II era processors and there
is no reason for hardware virtualizations to emulate these quirks.
Dimitry Andric [Thu, 21 May 2015 17:40:53 +0000 (17:40 +0000)]
Ensure that the static structs emitted by the MODULE_DEPEND() and
MODULE_VERSION() macros don't end up as .comm symbols, when all the
version fields are zero.
Normally, such symbols will end up in .bss, but for kernel module
version objects, this can lead to "garbage" version numbers.
Fix this by instructing the compiler to always put these structs in the
.data segment instead.
This is based on the patch sent by Alexander Fedorov with the following
fixes/improvements:
- Better error handling;
- Clock is derived from PLL6 (obtained from netbsd);
- No more unnecessary busy loops on interrupt handler;
- style(9) fixes and code cleanup.
I also want to thanks Martin Galvan who has sent an alternative
implementation with some interesting fixes.
Tested on CubieBoard2, Banana-Pi (thanks to netgate!) and Cubieboard1
(Pratik Singhal).
This is intended to pave the way for the upcoming GSoC work (and make
easier the build of images for the supported boards).
PR: 196081
Submitted by: Alexander Fedorov <alexander.fedorov@rtlservice.com>
John Baldwin [Thu, 21 May 2015 16:43:26 +0000 (16:43 +0000)]
Revert r282971. It depends on condvar consumers not destroying condvars
until all threads sleeping on a condvar have resumed execution after being
awakened. However, there are cases where that guarantee is very hard to
provide.
Pedro F. Giffuni [Thu, 21 May 2015 15:16:18 +0000 (15:16 +0000)]
ddb: finish converting boolean values.
The replacement started at r283088 was necessarily incomplete without
replacing boolean_t with bool. This also involved cleaning some type
mismatches and ansifying old C function declarations.
The RTC initialization values are based on the SoC which can be
determined at runtime so there's no need to set the values in
each DTS.
Tested on YYHD18 (aml8726-m3), VSATV102 (aml8726-m6), and
ODROIDC1 (aml8726-m8b).
Differential Revision: https://reviews.freebsd.org/D2588
Submitted by: John Wehle
Enji Cooper [Thu, 21 May 2015 05:06:08 +0000 (05:06 +0000)]
Import proposed fix from upstream for
atf-sh/atf_check_test:flush_stdout_on_timeout
Many thanks for jmmv for the fix!
MFC after: 1 week
PR: 197060
Original commit message:
From 0e546407567ea858e261e72f75c5ed61e07d0ddf Mon Sep 17 00:00:00 2001
From: Julio Merino <jmmv@google.com>
Date: Tue, 17 Feb 2015 18:10:11 -0500
Subject: [PATCH] Fix atf-sh/atf_check_test:flush_stdout_on_death
The test atf-sh/atf_check_test:flush_stdout_on_timeout was flaky as it
was playing solely with time. Fix this by making the test more robust
and rename it while we are at it: there is nothing left about "timeouts"
in this test, considering that ATF itself does not enforce deadlines
any longer.
Enji Cooper [Thu, 21 May 2015 05:02:08 +0000 (05:02 +0000)]
_Really_, _Really_ fix buildworld by moving the conditionals down, fixing some
typos, and fixing the dependency when MK_LIBCPLUSPLUS != no
- `:D` operator evaluation is immediate, i.e. like .if defined(..). So r283159
was in effect a no-op commit.
- Fix dependency in MK_LIBCPLUSPLUS case in two ways:
-- lib/libc++ was the wrong dependency. It should have been libcxxrt.
-- lib/libc++ was missing __L, so again it was depending on the directory, not
the relevant .PHONY target.
Tested with: make tinderbox (amd64, arm, sparc64) and JFLAG=-j16
In collaboration with: bdrewery, imp, peter
BIG pointyhat to: ngie (for trying to commit things at 6am while staying up all
night working on other tasks)
Peter Grehan [Thu, 21 May 2015 04:19:22 +0000 (04:19 +0000)]
Temporarily revert r282922 which bumped the max descriptors.
While there is no issued with the number of descriptors in
a virtio indirect descriptor, it's a guest's choice as to
whether indirect descriptors are used. For the case where
they aren't, the virtio block ring size is still 64 which
is less than the now reported max_segs of 67. This results
in an assertion in recent Linux guests even though it was
benign since they were using indirect descs.
The intertwined relationship between virtio ring size,
max seg size and blockif queue size will be addressed
in an upcoming commit, at which point the max descriptors
will again be bumped up to 67.
From 0e546407567ea858e261e72f75c5ed61e07d0ddf Mon Sep 17 00:00:00 2001
From: Julio Merino <jmmv@google.com>
Date: Tue, 17 Feb 2015 18:10:11 -0500
Subject: [PATCH] Fix atf-sh/atf_check_test:flush_stdout_on_death
The test atf-sh/atf_check_test:flush_stdout_on_timeout was flaky as it
was playing solely with time. Fix this by making the test more robust
and rename it while we are at it: there is nothing left about "timeouts"
in this test, considering that ATF itself does not enforce deadlines
any longer.
Do grammar fix in the comment to record the right commit message for
r283162.
Fix a cosmetic issue with vm_page_alloc() calling vm_page_free_toq()
with the page not completely satisfying vm_page_free() assertions.
The page is not owned by the object, since insertion failed. But
besides m->object reset to NULL, we should also set VPO_UNMANAGED flag
for consistency.
Reported by: pho
Reviewed by: alc
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Enji Cooper [Wed, 20 May 2015 13:04:00 +0000 (13:04 +0000)]
cddl/lib/libctf isn't always a requirement for lib/libproc; its use is dependent
on MK_CTF != "no". Use the other divined value instead of reinventing the wheel
Enji Cooper [Wed, 20 May 2015 12:26:09 +0000 (12:26 +0000)]
Build cddl/{sbin,usr.bin,usr.sbin} in parallel as all of the applications are
freestanding (they require libraries build via make libraries in buildworld)
In the reply to SADB_X_SPDGET message use the same sequence number that
was in the request. Some IKE deamons expect it will the same. Linux and
NetBSD also follow this behaviour.
D2432 (Patch for Amlogic single core PIC) moved the SMP option out
of AML8726 and into board specific config files since some boards
(e.g. YYHD18) use the aml8726-m3 which only have a single core.
r283057 applied most of D2432, however while it removed SMP from
AML8726, it missed adding the SMP option to the board specific
config files.
Differential Revision: https://reviews.freebsd.org/D2589
Submitted by: John Wehle
Warner Losh [Tue, 19 May 2015 21:16:53 +0000 (21:16 +0000)]
Re-select the SD card before getting the SD status. On a couple Atmel
boards, this prevents some error messages during enumeration and also
gives us the correct erase block size. They appear to be harmless
elsewhere.
# Note: we treat too many commands as 'can't fail' if they don't work
# after a couple of retries. We need to fix that, but not today...
Warner Losh [Tue, 19 May 2015 21:16:51 +0000 (21:16 +0000)]
Add NFS server to mix (for easier, in-place updates). Move to
partition 2 for root (since partition 1 is reserved for FAT
files the Atmel ROMs can load).
John Baldwin [Tue, 19 May 2015 19:15:19 +0000 (19:15 +0000)]
Fix two bugs that could result in PMC sampling effectively stopping.
In both cases, the the effect of the bug was that a very small positive
number was written to the counter. This means that a large number of
events needed to occur before the next sampling interrupt would trigger.
Even with very frequently occurring events like clock cycles wrapping all
the way around could take a long time. Both bugs occurred when updating
the saved reload count for an outgoing thread on a context switch.
First, the counter-independent code compares the current reload count
against the count set when the thread switched in and generates a delta
to apply to the saved count. If this delta causes the reload counter
to go negative, it would add a full reload interval to wrap it around to
a positive value. The fix is to add the full reload interval if the
resulting counter is zero.
Second, occasionally the raw counter value read during a context switch
has actually wrapped, but an interrupt has not yet triggered. In this
case the existing logic would return a very large reload count (e.g.
2^48 - 2 if the counter had overflowed by a count of 2). This was seen
both for fixed-function and programmable counters on an E5-2643.
Workaround this case by returning a reload count of zero.
Bring back support for checking tables via "ipfw -n".
Currently we have different table key types which can easily interfere
with each other (numbers and IPv4 address, interface names and hostnames,
flows and hostnames/addresses).
This conflicts are solved by [auto-]creating _typed_ tables, so after
table is created, only keys of given type can be inserted to that table.
ipfw(8) consults with kernel about key/value type for particular table so
it knows key/value interpretation.
However, we have 2 cases (adding entries to non-existing table and
parsing configuration file via `ipfw -n`) when kernel is unable to
provide us table info we need. Fix the latter case by partially importing
old `table_fill_xentry()` parse function responsible for guessing key type.
Alan Somers [Tue, 19 May 2015 16:23:47 +0000 (16:23 +0000)]
Properly null-terminate strings in a kernel dump header. A version string
longer than 192 bytes will cause the version field of a dump header to
overflow. strncpy doesn't null terminate it, so savecore will print a
corrupted info file. Using strlcpy fixes the bug.
Gleb Smirnoff [Tue, 19 May 2015 14:04:21 +0000 (14:04 +0000)]
Use MTX_SYSINIT() instead of mtx_init() to separate mutex initialization
from associated structures initialization. The mutexes are global, while
the structures are per-vnet.
Gleb Smirnoff [Tue, 19 May 2015 14:02:40 +0000 (14:02 +0000)]
During module unload unlock rules before destroying UMA zones, which
may sleep in uma_drain(). It is safe to unlock here, since we are already
dehooked from pfil(9) and all pf threads had quit.
Read GEOM_UNCOMPRESS metadata using several requests that fit into
MAXPHYS. For large compressed images the metadata size can be bigger
than MAXPHYS and this triggers KASSERT in g_read_data().
Also use g_free() to free memory allocated by g_read_data().
Teach key_expire() send SADB_EXPIRE message with the SADB_EXT_LIFETIME_HARD
extension header type. The key_flush_sad() now will send SADB_EXPIRE
message when HARD lifetime expires. This is required by RFC 2367 and some
keying daemons rely on these messages. HARD lifetime messages have
precedence over SOFT lifetime messages, so now they will be checked first.
Also now SADB_EXPIRE messages will be send even the SA has not been used,
because keying daemons might want to rekey such SA.
Adrian Chadd [Tue, 19 May 2015 05:41:00 +0000 (05:41 +0000)]
QCA955x / AP135 platform tidyups.
* Change mips24k -> mips74k for hwpmc, but leave it disabled for now.
* don't build pci by default.
* build pci and qca955x_pci for AP135, as theres a PCIe NIC.
* don't build a hwpmc module, it doesn't really work out well
for the mips boards at the moment.
* add ipfw and DEFAULT_TO_ACCEPT.
Adrian Chadd [Tue, 19 May 2015 05:31:58 +0000 (05:31 +0000)]
Add initial support for the QCA955x PCIe host controller.
The QCA955x looks a lot like the AR724x PCIe controller, except it
supports two root complexes. Unfortunately I only have one, so
although this code has started down the path of supporting more than
one, it's definitely not yet ready.
Tested:
* AP135 board (QCA9558 SoC), with the 11ac NIC swapped for an AR9380
PCIe NIC.
Notes:
* Yes, this driver isn't very pretty. I decided to commit what I have
versus holding onto something that isn't yet finished. It is enough
to bring up the above NIC and interrupt routing works, so it's a good
start.
* However, yes, the DDR flush routine hooks need to be fixed up.
I don't think I'm firing the right one at the moment.