Lawrence Stewart [Mon, 17 Jun 2013 09:49:07 +0000 (09:49 +0000)]
The fix committed in r250951 replaced the reported panic with a deadlock... gold
star for me. EVENTHANDLER_DEREGISTER() attempts to acquire the lock which is
held by the event handler framework while executing event handler functions,
leading to deadlock.
Move EVENTHANDLER_DEREGISTER() to alq_load_handler() and thus deregister the ALQ
shutdown_pre_sync handler at module unload time, which takes care of the
originally reported panic and fixes the deadlock introduced in r250951.
Reported by: Luiz Otavio O Souza
MFC after: 3 days
X-MFC with: 250951
Scott Long [Mon, 17 Jun 2013 08:57:09 +0000 (08:57 +0000)]
Add infrastructure for doing compatibility shims, as has been sorely
needed for the last 10 years. Far too much of the internal API is
exposed, and every small adjustment causes applications to stop working.
To kick this off, bump the API version to 0x17 as should have been done
with r246713, but add shims to compensate. Thanks to the shims, there
should be no visible change in application behavior.
I have plans to do a significant overhaul of the API to harnen it for
the future, but until then, I welcome others to add shims for older
versions of the API.
Jeff Roberson [Mon, 17 Jun 2013 03:43:47 +0000 (03:43 +0000)]
- Add a new UMA API: uma_zcache_create(). This makes a zone without any
backing memory that is only a container for per-cpu caches of arbitrary
pointer items. These zones have no kegs.
- Convert the regular keg based allocator to use the new import/release
functions.
- Move some stats to be atomics since they would require excessive zone
locking/unlocking with the new import/release paradigm. Make
zone_free_item simpler now that callers can manage more stats.
- Check for these cache-only zones in the public APIs and debugging
code by checking zone_first_keg() against NULL.
Pedro F. Giffuni [Sun, 16 Jun 2013 16:10:45 +0000 (16:10 +0000)]
Re-sort ext2fs headers to make things easier to find.
In the ext2fs driver we have a mixture of headers:
- The ext2_ prefixed headers have strong influence from NetBSD
and are carry specific ext2/3/4 information.
- The unprefixed headers are inspired on UFS and carry implementation
specific information.
Do some small adjustments so that the information is easier to
find coming from either UFS or the NetBSD implementation.
Ed Schouten [Sun, 16 Jun 2013 10:48:47 +0000 (10:48 +0000)]
Move _Atomic() into <sys/cdefs.h>.
That way _Atomic() is defined next to all the other C11 keywords for
which we provide compatibility for pre-C11 compilers. While there, fix
the definition to place "volatile" at the end. Otherwise pointer types
will become "volatile T *" instead of "T * volatile".
Ed Schouten [Sun, 16 Jun 2013 09:30:35 +0000 (09:30 +0000)]
Change callout use counter to use C11 atomics.
In order to get some coverage of C11 atomics in kernelspace, switch at
least one piece of code in kernelspace to use C11 atomics instead of
<machine/atomic.h>.
While there, slightly improve the code by adding an assertion to prevent
the use count from going negative.
Ed Schouten [Sat, 15 Jun 2013 22:17:59 +0000 (22:17 +0000)]
Let hastd use C11 atomics.
C11 atomics now work on all the architectures. Have at least a single
piece of software in our base system that uses C11 atomics. This
somewhat makes it less likely that we break it because of LLVM imports,
etc.
Alexander Motin [Sat, 15 Jun 2013 12:46:38 +0000 (12:46 +0000)]
Restore use of polling mode for disk cache flush in case of kernel panic.
While I am not sure that any extra hardware access is a good idea after
panic, that is an existing behaviour that should better work correctly.
Lawrence Stewart [Sat, 15 Jun 2013 10:38:31 +0000 (10:38 +0000)]
Now that the necessary infrastructure is in place to ensure hhook points which
register after a khelp module will get hooked, move khelp module initialisation
to the earlier SI_SUB_KLD stage.
Lawrence Stewart [Sat, 15 Jun 2013 10:08:34 +0000 (10:08 +0000)]
Move hhook's per-vnet initialisation to an earlier SYSINIT SI_SUB stage to
ensure all per-vnet related hhook initialisation is completed prior to any
virtualised hhook points attempting registration.
vnet_register_sysinit() requires that a stage later than SI_SUB_VNET be chosen.
There are no per-vnet initialisors in the source tree at this time which run
earlier than SI_SUB_INIT_IF. A quick audit of non-virtualised SYSINITs indicates
there are no subsystems pre SI_SUB_MBUF that would likely be interested in
registering a virtualised hhook point.
Settle on SI_SUB_MBUF as hhook's per-vnet initialisation stage as it's the first
overtly network-related initilisation stage to run after SI_SUB_VNET. If a
subsystem that initialises earlier than SI_SUB_MBUF ends up wanting to register
virtualised hhook points in future, hhook's use of SI_SUB_MBUF will need to be
revisited and would probably warrant creating a dedicated SI_SUB_HHOOK which
runs immediately after SI_SUB_VNET.
Ed Schouten [Sat, 15 Jun 2013 09:42:43 +0000 (09:42 +0000)]
Pull in r184040 from upstream clang trunk:
Emit native implementations of atomic operations on FreeBSD/armv6.
Just like on Linux, FreeBSD/armv6 assumes the system supports
ldrex/strex unconditionally. It is also used by the kernel. We can
therefore enable support for it, like we do on Linux.
While there, change one of the unit tests to explicitly test against
armv5 instead of armv7, as it actually tests whether libcalls are
emitted.
Ed Schouten [Sat, 15 Jun 2013 08:23:53 +0000 (08:23 +0000)]
Remove conflicting macros from SPARC64's atomic(9) header.
The atomic_load() and atomic_store() macros conflict with the equally
named macros from <stdatomic.h>. Remove them, as they are only used to
implement functions that are not present on any of the other
architectures.
Ed Schouten [Sat, 15 Jun 2013 08:15:22 +0000 (08:15 +0000)]
Make support for atomics on ARM complete.
Provide both __sync_*-style and __atomic_*-style functions that perform
the atomic operations on ARMv5 by using Restartable Atomic Sequences.
While there, clean up some pieces of code where it's sufficient to use
regular uint32_t to store register contents and don't need full reg_t's.
Also sync this back to the MIPS code.
Lawrence Stewart [Sat, 15 Jun 2013 05:57:29 +0000 (05:57 +0000)]
Add a private KPI between hhook and khelp that allows khelp modules to insert
hook functions into hhook points which register after the modules were loaded -
potentially useful during boot or if hhook points are dynamically registered.
Justin T. Gibbs [Sat, 15 Jun 2013 04:51:31 +0000 (04:51 +0000)]
Properly track the different reasons new I/O is temporarily disabled, and
only re-enable I/O when all reasons have cleared.
sys/dev/xen/blkfront/block.h:
In the block front driver softc, replace the boolean
XBDF_FROZEN flag with a count of commands and driver global
issues that freeze the I/O queue. So long xbd_qfrozen_cnt
is non-zero, I/O is halted.
Add flags to xbd_flags for tracking grant table entry and
free command resource shortages. Each of these classes can
increment xbd_qfrozen_cnt at most once.
Add a command flag (XBDCF_ASYNC_MAPPING) that is set whenever
the initial mapping attempt of a command fails with EINPROGRESS.
sys/dev/xen/blkfront/blkfront.c:
In xbd_queue_cb(), use new XBDCF_ASYNC_MAPPING flag to definitively
know if an async bus dmamap load has occurred.
Add xbd_freeze() and xbd_thaw() helper methods for managing
xbd_qfrozen_cnt and use them to implement all queue freezing logic.
Add missing "thaw" to restart I/O processing once grant references
become available.
Lawrence Stewart [Sat, 15 Jun 2013 04:03:40 +0000 (04:03 +0000)]
Internalise handling of virtualised hook points inside
hhook_{add|remove}_hook_lookup() so that khelp (and other potential API
consumers) do not have to care when they attempt to (un)hook a particular hook
point identified by id and type.
Justin T. Gibbs [Fri, 14 Jun 2013 23:43:44 +0000 (23:43 +0000)]
Upgrade Xen interface headers to Xen 4.2.1.
Move FreeBSD from interface version 0x00030204 to 0x00030208.
Updates are required to our grant table implementation before we
can bump this further.
sys/xen/hvm.h:
Replace the implementation of hvm_get_parameter(), formerly located
in sys/xen/interface/hvm/params.h. Linux has a similar file which
primarily stores this function.
sys/xen/xenstore/xenstore.c:
Include new xen/hvm.h header file to get hvm_get_parameter().
sys/amd64/include/xen/xen-os.h:
sys/i386/include/xen/xen-os.h:
Correctly protect function definition and variables from being
included into assembly files in xen-os.h
Xen memory barriers are now prefixed with "xen_" to avoid conflicts
with OS native primatives. Define Xen memory barriers in terms of
the native FreeBSD primatives.
Dimitry Andric [Fri, 14 Jun 2013 21:14:36 +0000 (21:14 +0000)]
Pull in r181620 from llvm trunk:
[ms-inline asm] Fix a crasher when we fail on a direct match.
The issue was that the MatchingInlineAsm and VariantID args to the
MatchInstructionImpl function weren't being set properly. Specifically, when
parsing intel syntax, the parser thought it was parsing inline assembly in the
at&t dialect; that will never be the case.
The crash was caused when the emitter tried to emit the instruction, but the
operands weren't set. When parsing inline assembly we only set the opcode, not
the operands, which is used to lookup the instruction descriptor.
rdar://13854391 and PR15945
Also, this commit reverts r176036. Now that we're correctly parsing the intel
syntax the pushad/popad don't match properly. I've reimplemented that fix using
a MnemonicAlias.
Pull in r183907 from llvm trunk:
X86: Make the cmov aliases work with intel syntax too.
These commits make a number of Intel-style inline assembly mnemonics
aliases (occurring in several ports) work properly, which could cause
assertions otherwise.
Devin Teske [Fri, 14 Jun 2013 19:35:30 +0000 (19:35 +0000)]
Add skeleton processing of selected packages in the order in which they
should be processed. De-select each package as we "process" (actual
processing still pending) each package so that if we get an error, we can
return to an accurate state of selected packages.
Lawrence Stewart [Fri, 14 Jun 2013 18:11:21 +0000 (18:11 +0000)]
Fix a major oversight in r251732 which causes non-VIMAGE kernels to trigger a
KASSERT during TCP hhook registration at boot. Virtualised hook points only
require extra housekeeping and sanity checking when "options VIMAGE" is present.
Justin T. Gibbs [Fri, 14 Jun 2013 17:00:58 +0000 (17:00 +0000)]
Improve debugger visibility into queuing functions by removing the macro
scheme for defining inline command queuing functions.
Prefer enums to #defines.
sys/dev/xen/blkfront/block.h
Replace inline function generation performed by the
XBDQ_COMMAND_QUEUE() macro with single instances of each
inline function (init, enqueue, dequeue, remove). This was
made possible by using queue indexes instead of bit flags
in the command structure, and passing the index enum as
an argument to the functions.
Improve panic/assert messages in the queue functions.
Combine queue data and stats into a single data structure
and declare an array of them instead of each queue individually.
Convert command flags, softc state, and softc flags to enums.
sys/dev/xen/blkfront/blkfront.c
Mechanical adjustments for new queue api.
Simon J. Gerraty [Fri, 14 Jun 2013 16:30:11 +0000 (16:30 +0000)]
Flag recursive make targets with .MAKE
so that job token pipe is passed to them.
To avoid surprising anyone, only add .MAKE to ${TGTS} when -n
has not been specified (at least for Makefile).
Simon J. Gerraty [Fri, 14 Jun 2013 16:24:07 +0000 (16:24 +0000)]
We cannot remove the _+_ trick, until old make is completely deprecated.
But we don't want to set it to + for bmake since it breaks make -N
which is used to supress the normal handling of targets marked with .MAKE
(which seems broken in fmake and might be why _+_ was introduced).
Add some comments to explain what's gong on.
Sergey Kandaurov [Fri, 14 Jun 2013 08:28:08 +0000 (08:28 +0000)]
Fix and improve filemon(4) example:
- remove return statements from void function [1]
- include missing header
- use O_CLOEXEC instead of separate fcntl() calls
Alexander Motin [Fri, 14 Jun 2013 08:26:58 +0000 (08:26 +0000)]
Improve firmware download status check. Previous check was insufficient for
ATA since it ignored transport errors like command timeouts, while for SCSI
it was just wrong.
Adrian Chadd [Fri, 14 Jun 2013 08:18:17 +0000 (08:18 +0000)]
Add in an initial WB225 (AR9485 + AR3012 BT) combo profile.
This hasn't yet been tested as unfortunately the AR3012 I have doesn't
have the "real" firmware on it; it shipped with the cut down HCI firmware
that only understands enough to accept a new firmware image.
Adrian Chadd [Fri, 14 Jun 2013 08:15:28 +0000 (08:15 +0000)]
The AR9300 HAL uses this config to program AR_PHY_SWITCH_COM_2 on AR9485
NICs which have bluetooth coexistence enabled.
The WB225 NIC has the common antenna switch configuration set to 0x0 which
disables all external switch bit setting. This obviously won't work when
doing coexistence.
This value is a magic value from the windows .inf files. It _looks_ right
but I haven't yet verified it - unfortunately my AR9285+AR3012 BT combo
has an earlier BT device which doesn't actually _have_ firmware on it.
So I have to fix ath3kfw to handle loading in firmware into the newer
NICs before I can finish testing this.
This may not hold true for CUS198, which is another custom AR9485 board.
Kevin Lo [Fri, 14 Jun 2013 05:36:47 +0000 (05:36 +0000)]
- Use the consistenly PHY-specific reset routine PHY_RESET() rather than
generic mii_phy_reset().
- Return the result of mii_mediachg() rather than blindly returning 0.
- on smsc(4), driver lock should be held to get current
mii_media_active/mii_media_status value.
Lawrence Stewart [Fri, 14 Jun 2013 04:10:34 +0000 (04:10 +0000)]
Add support for non-virtualised hhook points, which are uniquely identified by
type and id, as compared to virtualised hook points which are now uniquely
identified by type, id and a vid (which for vimage is the pointer to the vnet
that the hhook resides in).
All hhook_head structs for both virtualised and non-virtualised hook points
coexist in hhook_head_list, and a separate list is maintained for hhook points
within each vnet to simplify some vimage-related housekeeping.
Adrian Chadd [Fri, 14 Jun 2013 03:42:10 +0000 (03:42 +0000)]
Initial AR9485/AR933x 1x1 LNA diversity work.
* Add the LNA configuration table entries for AR933x/AR9485
* Add a chip-dependent LNA signal level delta in the startup path
* Add a TODO list for the stuff I haven't yet ported over but
I haven't.
Justin T. Gibbs [Fri, 14 Jun 2013 03:31:11 +0000 (03:31 +0000)]
sys/dev/xen/netfront/netfront.c:
In netif_free(), call ifmedia_removeall() after ether_ifdetach()
so that bpf listeners are detached, any link state processing
is completed, and there is no chance for external reference to media
information.
Lawrence Stewart [Fri, 14 Jun 2013 02:25:40 +0000 (02:25 +0000)]
Fix a potential NULL-pointer dereference that would trigger if the hhook
registration site did not provide storage for a copy of the hhook_head struct.
Marius Strobl [Thu, 13 Jun 2013 22:13:41 +0000 (22:13 +0000)]
All of Oxford/PLX OX16PCI954, OXm16PCI954 and OXu16PCI954 share the
exact same (subsystem) device and vendor IDs. However, the reference
design for the OXu16PCI954 uses a 14.7456 MHz clock (as does the EXSYS
EX-41098-2 equipped with these), while at least the OX16PCI954 defaults
to a 1.8432 MHz one. According to the datasheets of these chips, the
only difference in PCI configuration space is that OXu16PCI954 have
a revision ID of 1 while the other two are at 0. So employ the latter
for determining the default clock rates of this family.
Note that one might think that the actual clock could be derived from
the Clock Prescaler Register (CPR) of these chips. Unfortunately, this
is not that case and its use and content are orthogonal to the frequency
of the crystal employed.
Tested with an EXSYS EX-41098-2, which identifies and attaches as:
pcib4@pci0:19:0:0: class=0x060400 card=0x02dd1014 chip=0x10801b21
rev=0x03 hdr=0x01
vendor = 'ASMedia Technology Inc.'
device = 'ASM1083/1085 PCIe to PCI Bridge'
class = bridge
subclass = PCI-PCI
puc0@pci0:20:4:0: class=0x070006 card=0x00001415 chip=0x95011415
rev=0x01 hdr=0x00
vendor = 'Oxford Semiconductor Ltd'
device = 'OX16PCI954 (Quad 16950 UART) function 0 (Uart)'
class = simple comms
subclass = UART
puc1@pci0:20:4:1: class=0x068000 card=0x00001415 chip=0x95111415
rev=0x01 hdr=0x00
vendor = 'Oxford Semiconductor Ltd'
device = 'OX16PCI954 (Quad 16950 UART) function 1 (8bit bus)'
class = bridge
puc2@pci0:20:8:0: class=0x070006 card=0x00001415 chip=0x95011415
rev=0x01 hdr=0x00
vendor = 'Oxford Semiconductor Ltd'
device = 'OX16PCI954 (Quad 16950 UART) function 0 (Uart)'
class = simple comms
subclass = UART
puc3@pci0:20:8:1: class=0x068000 card=0x00001415 chip=0x95111415
rev=0x01 hdr=0x00
vendor = 'Oxford Semiconductor Ltd'
device = 'OX16PCI954 (Quad 16950 UART) function 1 (8bit bus)'
class = bridge
pci20: <ACPI PCI bus> on pcib4
puc0: <Oxford Semiconductor OX16PCI954 UARTs> port 0x5000-0x501f,
0x5020-0x503f mem 0xc6000000-0xc6000fff,0xc6001000-0xc6001fff irq 16 at
device 4.0 on pci20
uart1: <16950 or compatible> at port 1 on puc0
uart2: <16950 or compatible> at port 2 on puc0
uart3: <16950 or compatible> at port 3 on puc0
uart4: <16950 or compatible> at port 4 on puc0
puc1: <Oxford Semiconductor OX9160/OX16PCI954 UARTs (function 1)> port
0x5040-0x505f,0x5060-0x507f mem 0xc6002000-0xc6002fff,0xc6003000-0xc6003fff
irq 16 at device 4.1 on pci20
puc2: <Oxford Semiconductor OX16PCI954 UARTs> port 0x5080-0x509f,
0x50a0-0x50bf mem 0xc6004000-0xc6004fff,0xc6005000-0xc6005fff irq 16 at
device 8.0 on pci20
uart5: <16950 or compatible> at port 1 on puc2
uart6: <16950 or compatible> at port 2 on puc2
uart7: <16950 or compatible> at port 3 on puc2
uart8: <16950 or compatible> at port 4 on puc2
puc3: <Oxford Semiconductor OX9160/OX16PCI954 UARTs (function 1)> port
0x50c0-0x50df,0x50e0-0x50ff mem 0xc6006000-0xc6006fff,0xc6007000-0xc6007fff
irq 16 at device 8.1 on pci20
Andrew Turner [Thu, 13 Jun 2013 21:31:33 +0000 (21:31 +0000)]
Fix the vfp code to work with the 16 register variants of the VFP unit. We
check which variant we are on, and if it is a VFPv3 or v4, and has 32
double registers we save these. This fixes VFP support on Raspberry Pi.
While here clean fmrx and fmxr up to use the register names from vfp.h
as opposed to the raw register names.
Jeff Roberson [Thu, 13 Jun 2013 21:05:38 +0000 (21:05 +0000)]
- Convert the slab free item list from a linked array of indices to a
bitmap using sys/bitset. This is much simpler, has lower space
overhead and is cheaper in most cases.
- Use a second bitmap for invariants asserts and improve the quality of
the asserts as well as the number of erroneous conditions that we will
catch.
- Drastically simplify sizing code. Special case refcnt zones since they
will be going away.
- Update stale comments.
Ed Schouten [Thu, 13 Jun 2013 18:46:49 +0000 (18:46 +0000)]
Add C11 atomic fallbacks for ARM.
Basically the situation is as follows:
- When using Clang + armv6, we should not need any intrinsics. It should
support it, even though due to a target misconfiguration it does not.
We should fix this in Clang.
- When using Clang + noarmv6, provide __atomic_* functions that disable
interrupts.
- When using GCC + armv6, we can provide __sync_* intrinsics, similar to
what we did for MIPS. As ARM and MIPS are quite similar, simply base
this implementation on the one I did for MIPS.
- When using GCC + noarmv6, disable the interrupts, like we do for
Clang.
This implementation still lacks functions for noarmv6 userspace. To be
done.
Ed Schouten [Thu, 13 Jun 2013 18:40:45 +0000 (18:40 +0000)]
Minor improvements to <stdatomic.h>.
- Define __SYNC_ATOMICS in case we're using the __sync_*() API. This is
not used by <stdatomic.h> itself, but may be useful for some of the
intrinsics code to determine whether it should build the
machine-dependent intrinsic functions.
- Make is_lock_free() work in kernelspace. For now, assume atomics in
kernelspace are always lock free. This is a quite reasonable
assumption, as we surely shouldn't implement the atomic fallbacks for
arbitrary sizes.
Fix two issues with the spin loops in the umtx(2) implementation.
- When looping, check for the pending suspension. Otherwise, other
usermode thread which races with the looping one, could try to
prevent the process from stopping or exiting.
- Add missed checks for the faults from casuword*(). The code is
structured in a way which makes the loops exit if the specified
address is invalid, since both fuword() and casuword() return -1 on
the fault. But if the address is mapped readonly, the typical value
read by fuword() is different from -1, while casuword() returns -1.
Absent the checks for casuword() faults, this is interpreted as the
race with other thread and causes non-interruptible spinning in the
kernel.
Reported and tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Gleb Smirnoff [Thu, 13 Jun 2013 06:07:19 +0000 (06:07 +0000)]
Improve locking strategy between keys hash and ID hash.
Before this change state creating sequence was:
1) lock wire key hash
2) link state's wire key
3) unlock wire key hash
4) lock stack key hash
5) link state's stack key
6) unlock stack key hash
7) lock ID hash
8) link into ID hash
9) unlock ID hash
What could happen here is that other thread finds the state via key
hash lookup after 6), locks ID hash and does some processing of the
state. When the thread creating state unblocks, it finds the state
it was inserting already non-virgin.
Now we perform proper interlocking between key hash locks and ID hash
lock:
1) lock wire & stack hashes
2) link state's keys
3) lock ID hash
4) unlock wire & stack hashes
5) link into ID hash
6) unlock ID hash
To achieve that, the following hacking was performed in pf_state_key_attach():
- Key hash mutex is marked with MTX_DUPOK.
- To avoid deadlock on 2 key hash mutexes, we lock them in order determined
by their address value.
- pf_state_key_attach() had a magic to reuse a > FIN_WAIT_2 state. It unlinked
the conflicting state synchronously. In theory this could require locking
a third key hash, which we can't do now.
Now we do not remove the state immediately, instead we leave this task to
the purge thread. To avoid conflicts in a short period before state is
purged, we push to the very end of the TAILQ.
- On success, before dropping key hash locks, pf_state_key_attach() locks
ID hash and returns.
Pedro F. Giffuni [Thu, 13 Jun 2013 03:23:24 +0000 (03:23 +0000)]
Relax some unnecessary unsigned type changes in ext2fs.
While the changes in r245820 are in line with the ext2 spec,
the code derived from UFS can use negative values so it is
better to relax some types to keep them as they were, and
somewhat more similar to UFS. While here clean some casts.
Some of the original types are still wrong and will require
more work.