ian [Wed, 21 Oct 2015 13:59:00 +0000 (13:59 +0000)]
Move arm_gic_bind() out of the #ifdef SMP block to fix compile errors in
the not-SMP case. This is safe because arm_irq_next_cpu() will return
the cpuid of the current/only core in the not-SMP case.
ngie [Wed, 21 Oct 2015 12:54:15 +0000 (12:54 +0000)]
Unbreak makefs -t cd9660 after r289687
buffer_head needs to be freed -- not buffer
Detected by jemalloc, i.e. running makefs failed the arena assert
because my copy of malloc on CURRENT is compiled with the default
!MALLOC_PRODUCTION asserts on
royger [Wed, 21 Oct 2015 10:44:07 +0000 (10:44 +0000)]
xen: Code cleanup and small bug fixes
xen/hypervisor.h:
- Remove unused helpers: MULTI_update_va_mapping, is_initial_xendomain,
is_running_on_xen
- Remove unused define CONFIG_X86_PAE
- Remove unused variable xen_start_info: note that it's used inpcifront
which is not built at all
- Remove forward declaration of HYPERVISOR_crash
xen/xen-os.h:
- Remove unused define CONFIG_X86_PAE
- Drop unused helpers: test_and_clear_bit, clear_bit,
force_evtchn_callback
- Implement a generic version (based on ofed/include/linux/bitops.h) of
set_bit and test_bit and prefix them by xen_ to avoid any use by other
code than Xen. Note that It would be worth to investigate a generic
implementation in FreeBSD.
- Replace barrier() by __compiler_membar()
- Replace cpu_relax() by cpu_spinwait(): it's exactly the same as rep;nop
= pause
xen/xen_intr.h:
- Move the prototype of xen_intr_handle_upcall in it: Use by all the
platform
x86/xen/xen_intr.c:
- Use BITSET* for the enabledbits: Avoid to use custom helpers
- test_bit/set_bit has been renamed to xen_test_bit/xen_set_bit
- Don't export the variable xen_intr_pcpu
dev/xen/blkback/blkback.c:
- Fix the string format when XBB_DEBUG is enabled: host_addr is typed
uint64_t
dev/xen/balloon/balloon.c:
- Remove set but not used variable
- Use the correct type for frame_list: xen_pfn_t represents the frame
number on any architecture
dev/xen/control/control.c:
- Return BUS_PROBE_WILDCARD in xs_probe: Returning 0 in a probe callback
means the driver can handle this device. If by any chance xenstore is the
first driver, every new device with the driver is unset will use
xenstore.
dev/xen/grant-table/grant_table.c:
- Remove unused cmpxchg
- Drop unused include opt_pmap.h: Doesn't exist on ARM64 and it doesn't
contain anything required for the code on x86
dev/xen/netfront/netfront.c:
- Use the correct type for rx_pfn_array: xen_pfn_t represents the frame
number on any architecture
dev/xen/netback/netback.c:
- Use the correct type for gmfn: xen_pfn_t represents the frame number on
any architecture
dev/xen/xenstore/xenstore.c:
- Return BUS_PROBE_WILDCARD in xctrl_probe: Returning 0 in a probe callback
means the driver can handle this device. If by any chance xenstore is the
first driver, every new device with the driver is unset will use xenstore.
Note that with the changes, x86/include/xen/xen-os.h doesn't contain anymore
arch-specific code. Although, a new series will add some helpers that differ
between x86 and ARM64, so I've kept the headers for now.
royger [Wed, 21 Oct 2015 10:04:35 +0000 (10:04 +0000)]
x86/xen: Consolidate xen-os.h in a single place
amd64 and i386 platform code contain very similar xen/xen-os.h
The only differences are:
- Functions/variables/types which were unused in i386/xen/xen-os.h:
* xen_xchg
* __xchg_dummy
* __xg
* __xchg
* atomic_t
* atomic_inc
* rdtscll
The functions/variables/types unused in xen-os.h can be dropped and there
is no more differences betwen amd64 and i386.
The new header is placed in x86/include/xen and each platform will have
dummy headers include x86/xen/*.h. This is to be able to include
machine/xen/*.h in the PV drivers.
adrian [Wed, 21 Oct 2015 01:41:18 +0000 (01:41 +0000)]
arge: don't do the rx fixup copy and just offset the mbuf by 2 bytes
The existing code meets the "alignment" requirement for the l3 payload
by offsetting the mbuf by uint64_t and then calling an rx fixup routine
to copy the frame backwards by 2 bytes. This DWORD aligns the
L3 payload so tcp, etc doesn't panic on unaligned access.
This is .. slow.
For arge MACs that support 1 byte TX/RX address alignment, we can do
the "other" hack: offset the RX address of the mbuf so the L3 payload
again is hopefully DWORD aligned.
This is much cheaper - since TX/RX is both 1 byte align ready (thanks
to the previous commit) there's no bounce buffering going on and there
is no rx fixup copying.
This gets bridging performance up from 180mbit/sec -> 410mbit/sec.
There's around 10% of CPU cycles spent in _bus_dmamap_sync(); I'll
investigate that later.
Tested:
* QCA955x SoC (AP135 reference board), bridging arge0/arge1
by programming the switch to have two vlangroups in dot1q mode:
bdrewery [Wed, 21 Oct 2015 00:25:18 +0000 (00:25 +0000)]
Fix building in a directory with SUBDIRs and SUBDIR_PARALLEL.
The SUBDIR_PARALLEL feature uses a .for dir in ${SUBDIR} loop. The old code
here for recursing was setting SUBDIR= as a make *argument*. The SUBDIR=
replacement was not actually handled until after the .for loop was unrolled.
This could be seen with a '.info ${SUBDIR} ${dir}' inside of the loop which
showed an empty ${SUBDIR} and a set ${dir}. Setting NO_SUBIDR= before calling
${MAKE} as an *environment* variable handles the case fine and is a more
proper mechanism for disabling subdir handling.
This could be seen with 'make -C tests/sys/kern -j15 SUBDIR_PARALLEL=yes'.
adrian [Tue, 20 Oct 2015 21:18:02 +0000 (21:18 +0000)]
AR8327: Fix up the ability to configure the vlangroup configuration for the CPU port
I messed up when doing the reset_vlans method - setting vid[0] = 1 here
was making it 'hidden' from configuration (as it needed ETHERSWITCH_VID_VALID
as well) and so there was no way to configure vlangroup0.
In per-port VLAN mode, vlangroup0 is for the CPU port (port0).
Now, it normally wouldn't really matter - the CPU port thus sees
all other ports. However there are two CPU ports on the AR8327 and
so port0 (arge0) was seeing all traffic on port6 (arge1).
If you thus tried to use arge1/port6 for anything (eg a WAN port)
in a bridge group then things would very upset very quickly.
Whilst here, add a comment to remind myself that yes, it'd be nice
if we could specify a boot-time switch config.
kib [Tue, 20 Oct 2015 20:38:20 +0000 (20:38 +0000)]
Trim spaces at end of line to record the proper commit message for
r289660:
Do not allow to execute ptrace(PT_TRACE_ME) when the process is
already traced.
Do not allow to execute ptrace(PT_TRACE_ME) when there is no parent
which can trace the process, i.e. when the parent is already init.
Note that after the PT_TRACE_ME request the process is unkillable and
non-continuable until a debugger is attached, or parent is killed, the
later clears P_TRACED state. Since init clearly would not debug the
caller, and cannot be killed, disallow creation of unkillable
processes.
Reviewed by: jhb, pho
Reported and tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Differential revision: https://reviews.freebsd.org/D3908
bdrewery [Tue, 20 Oct 2015 20:37:00 +0000 (20:37 +0000)]
Improve safety of caching from r289659 by only importing of none of the
variables are already set. This should cover odd cases such as the
COMPILER_TYPE override in lib/csu/powerpc64.
jmmv [Tue, 20 Oct 2015 20:35:34 +0000 (20:35 +0000)]
Handle lib32 files during delete-old* when MK_LIB32=no.
Extend OptionalObsoleteFiles.inc to delete all lib32 files when MK_LIB32 is
set to no on a system that previously had lib32 libraries installed.
Also, to prevent "make delete-old-dirs" from always deleting lib32 directories
after an installworld, move the lib32 subtree to its own mtree file that only
gets applied when MK_LIB32=yes.
Test: Ran "make delete-old" and "make delete-old-libs" on a system that never
had MK_LIB32 enabled, and on a system where MK_LIB32 was enabled and later
disabled. Did this both on amd64 and powerpc64.
kib [Tue, 20 Oct 2015 20:29:21 +0000 (20:29 +0000)]
Mark struct thread zone as type-stable.
When establishing the locking state for several lock types (including
blockable mutexes and sx) failed, locking primitives try to spin while
the owner thread is running. The spinning loop performs the test for
running condition by dereferencing the owner->td_state field of the
owner thread. If the owner thread exited while spinner was put off
the processor, it is harmless to access reused struct thread owner,
since in some near future the current processor would notice the owner
change and make appropriate progress. But it could be that the page
which carried the freed struct thread was unmapped, then we fault
(this cannot happen on amd64).
For now, disallowing free of the struct thread seems to be good
enough, and tests which create a lot of threads once, did not
demonstrated regressions.
Reviewed by: jhb, pho
Reported and tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Differential revision: https://reviews.freebsd.org/D3908
bdrewery [Tue, 20 Oct 2015 20:15:25 +0000 (20:15 +0000)]
Pass COMPILER_TYPE and COMPILER_VERSION to sub-makes to avoid redundant
lookups.
This uses a special variable name based on a hash of ${CC}, ${PATH}, and
${MACHINE} to ensure that a cached value is not used if any of these
values changes to use a new compiler.
Before this there were 34,620 fork/exec from bsd.compiler.mk during a buildworld.
After this there are 608. More improvement is needed to cache a value from
the top-level before descending into subdirs in the various build phases.
dumbbell [Tue, 20 Oct 2015 19:52:59 +0000 (19:52 +0000)]
iicbus: Use device_delete_children() instead of explicit child removal
If the bus is detached and deleted by a call to device_delete_child() or
device_delete_children() on a device higher in the tree, I²C children
were already detached and deleted. So the device_t pointer stored in sc
points to freed memory: we must not try to delete it again.
By using device_delete_children(), we let subr_bus.c figure out if there
are children to take care of.
While here, make sure iicbus_detach() and iicoc_detach() call
device_delete_children() too, to be safe.
It was possible for a synchronous update of the RX index in the error
case to get ahead of the asynchronous RX index update in the normal
case. Change the RX processing to preserve an RX completion order.
There were two error cases. First, if a buffer is not present to
receive data, there would be no queue entry to preserve the RX
completion order. Instead of dropping the RX frame, leave the RX frame
in the ring. Schedule RX processing when RX entries are enqueued, in
case there are RX frames waiting in the ring to be received.
Second, if a buffer is too small to receive data, drop the frame in the
ring, mark the RX entry as done, and indicate the error in the RX entry
length. Check for a negative length in the receive callback in
ntb_netdev, and count occurrences as rx_length_errors.
Authored by: Allen Hubbe
Obtained from: Linux (Dual BSD/GPL driver)
Sponsored by: EMC / Isilon Storage Division
ian [Tue, 20 Oct 2015 15:15:30 +0000 (15:15 +0000)]
Uncomment some rather important code that was commented out for benchmarking.
Normally this routine is supposed to loop until the PIC returns a "no more
interrupts pending" indication. I had commented that out to do just one
interrupt per invokation to do some timing tests.
hselasky [Tue, 20 Oct 2015 09:13:35 +0000 (09:13 +0000)]
Merge LinuxKPI changes from DragonflyBSD:
- Remove redundant NBLONG macro and use BIT_WORD()
and BIT_MASK() instead.
- Correctly define BIT_MASK() according to Linux and
update all users of this macro.
- Add missing GENMASK() macro.
- Remove all comments deriving from Linux.
cem [Tue, 20 Oct 2015 01:54:34 +0000 (01:54 +0000)]
if_ntb: Fix typo in qp_link_work to match Linux
Throw away the result of the peer SPAD read. The peer will write our
local SPAD and we need to keep the locally read SPAD value to check if
the remote side is up.
cem [Tue, 20 Oct 2015 01:54:16 +0000 (01:54 +0000)]
if_ntb: MFV 2849b5d7: Reset transport QP link stats on down
Reset the link stats when the link goes down. In particular, the TX and
RX index and count must be reset, or else the TX side will be sending
packets to the RX side where the RX side is not expecting them. Reset
all the stats, to be consistent.
Authored by: Allen Hubbe
Obtained from: Linux (Dual BSD/GPL driver)
Sponsored by: EMC / Isilon Storage Division
cem [Tue, 20 Oct 2015 01:46:14 +0000 (01:46 +0000)]
NTB: MFV 5ae0beb6: Enable link for Intel root port mode in probe
We skip actually bringing up Rootport/Transparent configurations, so
most of this doesn't apply. Original Linux commit log:
Link training should be enabled in the driver probe for root port mode.
We should not have to wait for transport to be loaded for this to
happen. Otherwise the ntb device will not show up on the transparent
bridge side of the link.
Authored by: Dave Jiang
Obtained from: Linux (Dual BSD/GPL driver)
Sponsored by: EMC / Isilon Storage Division
bdrewery [Mon, 19 Oct 2015 23:34:35 +0000 (23:34 +0000)]
Replace all of the duplicated logic for recursing into a subdir with one
implementation. It is duplicated at run-time but is more easily
maintainable now.
ian [Mon, 19 Oct 2015 19:18:02 +0000 (19:18 +0000)]
Set the correct values in the arm aux control register, based on chip type.
The bits in the aux control register vary based on the processor type. In
the past we've always just set the 'smp' and "broadcast tlb/cache ops' bits,
which worked fine for the first few SoCs we supported. Now that we support
most of the cortex-a series processors, it's important to get the right bits
set based on the processor type.
hselasky [Mon, 19 Oct 2015 16:03:08 +0000 (16:03 +0000)]
The returned value from vm_fault_disable_pagefaults() must be stored
and passed to vm_fault_enable_pagefaults(). Else possible recursion on
the state can be lost.
royger [Mon, 19 Oct 2015 14:47:37 +0000 (14:47 +0000)]
xen-netfront: fix netfront create_dev error path
The failure path for allocating rx grant refs should not try to free tx
grant refs because tx grant refs were allocated after that. Also fix the
error path for xen_net_read_mac.
Submitted by: Wei Liu <wei.liu2@citrix.com>
Reviewed by: royger
Differential Revision: https://reviews.freebsd.org/D3891
Sponsored by: Citrix Systems R&D
royger [Mon, 19 Oct 2015 14:20:06 +0000 (14:20 +0000)]
xen-netfront: purge page flipping support
Currently neither Linux nor FreeBSD netback supports page flipping. NetBSD
still supports that. It is not sure how many people actually use page
flipping, but page flipping is supposed to be slower than copying nowadays.
It will also shatter frontend / backend address space.
Overall this feature is more of a burden than a benefit.
Submitted by: Wei Liu <wei.liu2@citrix.com>
Reviewed by: royger
Differential Revision: https://reviews.freebsd.org/D3888
Sponsored by: Citrix Systems R&D
andrew [Mon, 19 Oct 2015 13:20:23 +0000 (13:20 +0000)]
Use 4 levels of page tables when enabling the MMU. This will allow us to
boot on an SoC that places physical memory at an address past where three
levels of page tables can access in an identity mapping.
Submitted by: Wojciech Macek <wma@semihalf.com>,
Patrick Wildt <patrick@bitrig.org>
Differential Revision: https://reviews.freebsd.org/D3885 (partial)
Differential Revision: https://reviews.freebsd.org/D3744