Michal Meloun [Fri, 19 Aug 2016 10:52:39 +0000 (10:52 +0000)]
INTRNG: Rework handling with resources. Partially revert r301453.
- Read interrupt properties at bus enumeration time and store
it into global mapping table.
- At bus_activate_resource() time, given mapping entry is resolved and
connected to real interrupt source. A copy of mapping entry is attached
to given resource.
- At bus_setup_intr() time, mapping entry stored in resource is used
for delivery of requested interrupt configuration.
- For MSI/MSIX interrupts, mapping entry is created within
pci_alloc_msi()/pci_alloc_msix() call.
- For legacy PCI interrupts, mapping entry must be created within
pcib_route_interrupt() by pcib driver itself.
Reviewed by: nwhitehorn, andrew
Differential Revision: https://reviews.freebsd.org/D7493
Warner Losh [Fri, 19 Aug 2016 04:30:29 +0000 (04:30 +0000)]
Improve the pattern matching so that internal *'s work, as well as
[set] notation. This fixes pattern matching for recently added drives
that would set the NCQ Trim being broken incorrectly.
PR: 210686 Tested-by: Tomoaki AOKI
MFC After: 3 days
Kirk McKusick [Fri, 19 Aug 2016 00:03:41 +0000 (00:03 +0000)]
Fsck_ufs was using an int rather than a ufs2_daddr_t to store the
alternate superblock location when given in the -b option. When int
is 32-bits, block numbers larger than 2^32 would get truncated. This
commit changes the storage fpr the alternate superblock location
to a ufs2_daddr_t.
Ryan Stone [Thu, 18 Aug 2016 22:59:10 +0000 (22:59 +0000)]
Fix unlocked access to ifnet address list
in_broadcast() was iterating over the ifnet address list without
first taking an IF_ADDR_RLOCK. This could cause a panic if a
concurrent operation modified the list.
Ryan Stone [Thu, 18 Aug 2016 22:59:00 +0000 (22:59 +0000)]
Don't iterate over the ifnet addr list in ip_output()
For almost every packet that is transmitted through ip_output(),
a call to in_broadcast() was made to decide if the destination
IP was a broadcast address. in_broadcast() iterates over the
ifnet's address to find a source IP matching the subnet of the
destination IP, and then checks if the IP is a broadcast in that
subnet.
This is completely redundant as we have already performed the
route lookup, so the source IP is already known. Just use that
address to directly check whether the destination IP is a
broadcast address or not.
Mark Johnston [Thu, 18 Aug 2016 17:27:58 +0000 (17:27 +0000)]
Add a SIGINFO handler for dtrace(1).
Have it print the contents of aggregations, if any. Otherwise, one needs to
kill the running script to view the collected data, or add code to
periodically print it.
Ed Maste [Thu, 18 Aug 2016 16:22:40 +0000 (16:22 +0000)]
vt: fix old keyboard release in CONS_SETKBD
On the first switch we previously released the newly allocated keyboard
instead of the old one. Keyboard state was very confused afterwards for
further keyboard switches.
Bjoern A. Zeeb [Thu, 18 Aug 2016 10:26:15 +0000 (10:26 +0000)]
Try to fix gcc compilation errors (which are right).
nat64_getlasthdr() returns an int, which can be -1 in case of error,
storing the result in an uint8_t and then comparing to < 0 is not
helpful. Do what is done in the rest of the code and make proto an
int here as well.
Add support for setting blocking and non-blocking mode on /dev/rdma_cm
by returning success on FIONBIO and FIOASYNC IOCTLs. The actual flags
handling is done by the kern_ioctl() function.
Pyun YongHyeon [Thu, 18 Aug 2016 06:29:07 +0000 (06:29 +0000)]
Introduce axge_rxfilter() which configures RX filtering and replace
axge_setmulti()/axge_setpromisc() with axge_rxfilter().
Multicast filter programming and promiscuous mode requires
access to a common RX configuration register so there is no need to
use separate functions with added complexity. axge_rxfilter() does
not read back AXGE_RCR register since accessing a register in USB
is too slow and we already have all knowledge of required
configuration. Rebuilding RX filter configuration is simpler and
faster than manipulating every bits after reading back the
register.
Note, axge_rxfilter() does not set RCR_IPE(IP header alignment on
32bit boundary) to disable extra padding bytes insertion. The
extra padding wastes ethernet to USB host bandwidth as well as
complicating RX handling logic. Current USB framework requires
copying RX frames to mbufs so there is no need to worry about
alignment. Previously axge_rx_frame() performed wrong bound check
due to the extra padding and it was broken when RX checksum
offloading is disabled. See added comment in axge_rx_frame () for
actual RX packet layout.
In axge_init(), disable WOL. It's meaningless to enable WOL in
normal operation.
In axge_rxeof(), use properly sized mbuf rather than blindly
allocating a mbuf cluster.
Use RX H/W checksum offloading only when administrator requested RX
checksum offloading. Previously it always used RX H/W checksum
offloading result regardless of RX checksum offloading state.
Separate L4 checksum offloading validation from L3 one and properly
set required offloading bits for each layer. This is to fix setting
L4 checksum offloading bits for L3 packets.
There are still lots of RX errors(probably RX FIFO overflows) under
moderate load. Users are strongly recommended to enable ethernet
flow control.
Pyun YongHyeon [Thu, 18 Aug 2016 05:07:02 +0000 (05:07 +0000)]
Switch to TX header format rather than directly manipulating header
structures. This simplifies mbuf copy operation to USB buffers as
well as improving readability. The controller supports Microsoft
LSOv1(aka TSO) but this change set does not include the support due
to copying overhead to USB buffers and large amount of memory waste.
Remove useless ZLP padding which seems to come from Linux. Required
bits the code tried to set was not copied into USB buffer so it had
no effect. Unlike Linux, FreeBSD USB stack automatically generates
ZLP so no explicit padding is required in driver.[1]
Micro-optimize updating IFCOUNTER_OPACKETS counter by moving it out
of TX loop since updating counter is not cheap operation as it did
long time ago and we already know how many number of packets were
queued after exiting the loop.
While here, fix a checksum offloading bug which will happen when
upper stack computes checksum while H/W checksum offloading is
active. The controller should be notified to not recompute the
checksum in this case.
Pyun YongHyeon [Thu, 18 Aug 2016 02:14:39 +0000 (02:14 +0000)]
Don't explicitly call MIIBUS_STATCHG() method handler. Link state
change should be handled by PHY driver. Some broken PHY H/Ws may
need that workaround but it seems axge(4) don't use such PHYs.
Toomas Soome [Thu, 18 Aug 2016 00:37:07 +0000 (00:37 +0000)]
Add SHA512, skein, large blocks support for loader zfs.
Updated sha512 from illumos.
Using skein from freebsd crypto tree.
Since loader itself is using 64MB memory for heap, updated zfsboot to
use same, and this also allows to support zfs large blocks.
Note, adding additional features does increate zfsboot code, therefore
this update does increase zfsboot code to 128k, also I have ported gptldr.S
update to zfsldr.S to support 64k+ code.
With this update, boot1.efi has almost reached the current limit of the size
set for it, so one of the future patches for boot1.efi will need to
increase the limit.
Currently known missing zfs features in boot loader are edonr and gzip support.
Reviewed by: delphij, imp
Approved by: imp (mentor)
Obtained from: sha256.c update and skein_zfs.c stub from illumos.
Differential Revision: https://reviews.freebsd.org/D7418
Jung-uk Kim [Wed, 17 Aug 2016 22:13:39 +0000 (22:13 +0000)]
Disable assembly sources when compiler/assembler cannot compile certain
instructions. For example, GCC 4.2.1 + binutils 2.17.50 does not support
AVX instructions.
Emmanuel Vadot [Wed, 17 Aug 2016 20:32:08 +0000 (20:32 +0000)]
Rename kernel config A10 into ALLWINNER_UP as it is intend to work with all
Allwinner Uniprocessor SoC.
As of now it works with A10 and A13 (and possibly R8 as it is the same as the A13).
Move files.a10 into a1o subdirectory as it should be.
Rename std.a10 into std.allwinner_up
This adds support for EARLY_PRINTF via the CFE console; the aim is to
provide a fix for the otherwise cyclic dependency between PMU discovery
and console printf/DELAY:
- We need to parse the bhnd(4) core table to determine the address (and
type) of the PMU/PLL registers and calculate the CPU clock frequency.
- The core table parsing code will emit a printf() if a parse error is
hit.
- Safely calling printf() without EARLY_PRINTF requires a working
DELAY+cninit, which means we need the PMU.
Errors in core table parsing shouldn't happen, but lack of EARLY_PRINTF
makes debugging more difficult.
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D7498
Teach netisr_get_cpuid() to limit a given value to supported by netisr.
Use netisr_get_cpuid() in netisr_select_cpuid() to limit cpuid value
returned by protocol to be sure that it is not greather than nws_count.
Emmanuel Vadot [Wed, 17 Aug 2016 13:09:31 +0000 (13:09 +0000)]
Only set pud settings if this is a pullup or pulldown configuration.
This removes the need to set the MMC pins with pullups in our DTS.
Thanks to jmcneill@ for spotting this.
Emmanuel Vadot [Wed, 17 Aug 2016 10:20:36 +0000 (10:20 +0000)]
a10_gpio_get_function now returns the whole function not only
GPIO_INPUT/GPIO_OUTPUT.
a10_gpio_get_pud now returns the whole pud not only PULLDOWN/PULLUP.
Add a10_gpio_get_drv to get the current drive strenght.
During fdt pin configure, avoid setting function/drive/pud if it's already in
the correct value.
Implement userspace gettimeofday(2) with HPET timecounter.
Right now, userspace (fast) gettimeofday(2) on x86 only works for
RDTSC. For older machines, like Core2, where RDTSC is not C2/C3
invariant, and which fall to HPET hardware, this means that the call
has both the penalty of the syscall and of the uncached hw behind the
QPI or PCIe connection to the sought bridge. Nothing can me done
against the access latency, but the syscall overhead can be removed.
System already provides mappable /dev/hpetX devices, which gives
straight access to the HPET registers page.
Add yet another algorithm to the x86 'vdso' timehands. Libc is updated
to handle both RDTSC and HPET. For HPET, the index of the hpet device
to mmap is passed from kernel to userspace, index might be changed and
libc invalidates its mapping as needed.
Remove cpu_fill_vdso_timehands() KPI, instead require that
timecounters which can be used from userspace, to provide
tc_fill_vdso_timehands{,32}() methods. Merge i386 and amd64
libc/<arch>/sys/__vdso_gettc.c into one source file in the new
libc/x86/sys location. __vdso_gettc() internal interface is changed
to move timecounter algorithm detection into the MD code.
Measurements show that RDTSC even with the syscall overhead is faster
than userspace HPET access. But still, userspace HPET is three-four
times faster than syscall HPET on several Core2 and SandyBridge
machines.
Tested by: Howard Su <howard0su@gmail.com>
Sponsored by: The FreeBSD Foundation
MFC after: 1 month
Differential revision: https://reviews.freebsd.org/D7473
By default, allow all to read the HPET registers pages. At the same
time, by, by default disallow writes to the mmaped HPET pages.
Intent is to allow userspace to use HPET as fast (i.e. no-syscall)
timecounter for gettimeofday(2). Unfortunately, the permission model
does not make it possible to safely unhide /dev/hpet in the jails even
if default mode is set to 0444, because untrusted jailed root may
change device permissions to writeable.
Sponsored by: The FreeBSD Foundation
MFC after: 3 weeks
Sepherosa Ziehau [Wed, 17 Aug 2016 05:02:18 +0000 (05:02 +0000)]
hyperv/storvsc: Deliver CAM_SEL_TIMEOUT upon SRB status error.
SRB status is set to 0x20 by the hypervisor, if the specified LUN is
unaccessible, and even worse the INQUIRY response will not be set by
the hypervisor at all under this situation. Additionally, SRB status
is 0x20 too, for TUR on an unaccessible LUN.
Deliver CAM_SEL_TIMEOUT to CAM upon SRB status errors as suggested by
Scott Long, other values seems improper.
This commit fixes the Hyper-V disk hotplug support.
Submitted by: Hongjiang Zhang <honzhan microsoft com>
MFC after: 3 days
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7521
Eric van Gyzen [Wed, 17 Aug 2016 01:24:34 +0000 (01:24 +0000)]
PCIe HotPlug: Detect bridges that are not really HotPlug capable
Some devices report that they have an MRL when they actually
do not. Since they always report that the MRL is open, child
devices would be ignored. Try to detect these devices and
ignore their claim of HotPlug support. Specifically,
if there is an open MRL but the Data Link Layer is active,
the MRL is not real.
Revert r303645 to re-enable HotPlug support for slots with
power controllers, since it works correctly in my testing.
Start the DLL state-change timer if Presence /or/ MRL state changes,
along with other conditions. Previously, we started the timer iff
Presence changed. If there is an MRL, it must be closed for power
to be turned on, so Presence is unlikely to change on an MRL-close event.
Add a printf() of interesting registers on HotPlug interrupts and
commands (one from erj@). These were very useful for debugging.
Guard them with bootverbose, since they're spam in normal operation.
In collaboration with: jhb
Reviewed by: jhb
MFC after: 1 day
Relnotes: yes (re-enable HotPlug support for slots with power controllers)
Sponsored by: Dell Inc.
Differential Revision: https://reviews.freebsd.org/D7509
Gleb Smirnoff [Tue, 16 Aug 2016 21:55:34 +0000 (21:55 +0000)]
We should not be allowing a timeout to reset when a drain is in progress on
it (either async or sync drain).
At this moment the only user of drain is TCP, but TCP wouldn't reschedule a
callout after it has drained it, since it drains only when a tcpcb is closed.
This for now the problem isn't observed.
Landon J. Fuller [Tue, 16 Aug 2016 21:32:05 +0000 (21:32 +0000)]
bhnd(4): Implement NVRAM support required for PMU bring-up.
- Added a generic bhnd_nvram_parser API, with support for the TLV format
used on WGT634U devices, the standard BCM NVRAM format used on most
modern devices, and the "board text file" format used on some hardware
to supply external NVRAM data at runtime (e.g. via an EFI variable).
- Extended the bhnd_bus_if and bhnd_nvram_if interfaces to support both
string-based and primitive data type variable access, required for
common behavior across both SPROM and NVRAM data sources.
- Extended the existing SPROM implementation to support the new
string-based NVRAM APIs.
- Added an abstract bhnd_nvram driver, implementing the bhnd_nvram_if
atop the bhnd_nvram_parser API.
- Added a CFE-based bhnd_nvram driver to provide read-only access to
NVRAM data on MIPS SoCs, pending implementation of a flash-aware
bhnd_nvram driver.
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D7489
Landon J. Fuller [Tue, 16 Aug 2016 21:20:05 +0000 (21:20 +0000)]
bhndb(4): Drop MIPS-incompatible __builtin_ctz dependency.
This replaces the bitfield representation of the bhndb register window
freelist with the bitstring API, eliminating a dependency on
(MIPS-unsupported) __builtin_ctz().
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D7495
Kirk McKusick [Tue, 16 Aug 2016 21:02:30 +0000 (21:02 +0000)]
Bug 211013 reports that a write error to a UFS filesystem running
with softupdates panics the kernel. The problem that has been pointed
out is that when there is a transient write error on certain metadata
blocks, specifically directory blocks (PAGEDEP), inode blocks
(INODEDEP), indirect pointer blocks (INDIRDEPS), and cylinder group
(BMSAFEMAP, but only when journaling is enabled), we get a panic
in one of the routines called by softdep_disk_io_initiation that
the I/O is "already started" when we retry the write.
These dependency types potentially need to do roll-backs when called
by softdep_disk_io_initiation before doing a write and then a
roll-forward when called by softdep_disk_write_complete after the
I/O completes. The panic happens when there is a transient error.
At the top of softdep_disk_write_complete we check to see if the
write had an error and if an error occurred we just return. This
return is correct most of the time because the main role of the routines
called by softdep_disk_write_complete is to process the now-completed
dependencies so that the next I/O steps can happen.
But for the four types listed above, they do not get to do their
rollback operations. This causes the panic when softdep_disk_io_initiation
gets called on the second attempt to do the write and the roll-back
routines find that the roll-backs have already been done. As an
aside I note that there is also the problem that the buffer will
have been unlocked and thus made visible to the filesystem and to
user applications with the roll-backs in place.
The way to resolve the problem is to add a flag to the routines called
by softdep_disk_write_complete for the four dependency types noted
that indicates whether the write was successful (WRITESUCCEEDED).
If the write does not succeed, they do just the roll-backs and then
return. If the write was successful they also do their usual
processing of the now-completed dependencies.
The fix was tested by selectively injecting write errors for buffers
holding dependencies of each of the four types noted above and then
verifying that the kernel no longer paniced and that following the
successful retry of the write that the filesystem could be unmounted
and successfully checked cleanly.
In UFS_BALLOC(), invalidate pages of indirect buffers on failed block
allocation unwinding.
Dandling buffers are released on UFS_BALLOC() failure to ensure that
later attempt to allocate blocks in close range do not find the blocks
with invalid content, since possible partial block allocations are
unwound. As such, it is not enough to just release the buffers, the
pages must also invalidated and removed from the vnode vm_object
queue. Otherwise the pages might be found later and used to
reconstruct indirect buffers when doing allocations at offset close to
the failure point, and their stale content compromise the filesystem
integrity.
Note that just marking the buffer as B_INVAL is not enough, B_NOCACHE
is required. To be sure, clear the B_CACHE flag as well. This
complements the r174973, which started releasing buffers.
Reported and tested by: pho
Reviewed by: mckusick
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
On unwind after failed block allocation in ffs_balloc_ufs{1,2}, assert
that recorded allocated blocks numbers match the physical block
numbers of dandling buffers which are released.
When finally freeing the blocks during unwind, assert that dandling
buffers where not re-allocated. They shouldn't, because the vnode lock
is owned exclusive.
Reviewed by: mckusick
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Kirk McKusick [Tue, 16 Aug 2016 17:07:48 +0000 (17:07 +0000)]
Add two new macros, SLIST_CONCAT and LIST_CONCAT. Note in both the
queue.h header file and in the queue.3 manual page that they are O(n)
so should be used only in low-usage paths with short lists (otherwise
an STAILQ or TAILQ should be used).
When block allocation fails in UFS_BALLOC(), and the volume does not
have SU enabled, there is no point in calling softdep_request_cleanup().
The call cannot produce free blocks, but we unecessarily lock ufsmount
and do inode block write. Usual point of not doing optimizations for
the corner case of the full volume is not applicable there, the work
is easily avoidable, and the addition CPU and disk io load do not lead
to succeeding retry.
Reviewed by: mckusick
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
In ffs_balloc_ufs{1,2} routines, assert that unwind records do not
overflow local arrays. This is not immediately obvious from the
static code inspection, due to retry logic.
Reviewed by: mckusick
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Randall Stewart [Tue, 16 Aug 2016 15:11:46 +0000 (15:11 +0000)]
Here we update the modular tcp to be able to switch to an
alternate TCP stack in other then the closed state (pre-listen/connect).
The idea is that *if* that is supported by the alternate stack, it
is asked if its ok to switch. If it approves the "handoff" then we
allow the switch to happen. Also the fini() function now gets a flag
to tell if you are switching away *or* the tcb is destroyed. The
init() call into the alternate stack is moved to the end so the
tcb is more fully formed before the init transpires.
Sponsored by: Netflix Inc.
Differential Revision: D6790