lwhsu [Fri, 31 May 2019 04:29:29 +0000 (04:29 +0000)]
Remove tests for the deprecated algorithms in r348206
The tests are failing because the return value and output have changed, but
before test code structure adjusted, removing these test cases help people
be able to focus on more important cases.
Discussed with: emaste
MFC with: r348206
Sponsored by: The FreeBSD Foundation
rmacklem [Fri, 31 May 2019 01:28:48 +0000 (01:28 +0000)]
Replace a single linked list with a hash table of lists.
mountd.c uses a single linked list of "struct exportlist" structures,
where there is one of these for each exported file system on the NFS server.
This list gets long if there are a large number of file systems exported and
the list must be searched for each line in the exports file(s) when
SIGHUP causes the exports file(s) to be reloaded.
A simple benchmark that traverses SLIST() elements and compares two 32bit
fields in the structure for equal (which is what the search is)
appears to take a couple of nsec. So, for a server with 72000 exported file
systems, this can take about 5sec during reload of the exports file(s).
By replacing the single linked list with a hash table with a target of
10 elements per list, the time should be reduced to less than 1msec.
Peter Errikson (who has a server with 72000+ exported file systems) ran
a test program using 5 hashes to see how they worked.
fnv_32_buf(fsid,..., 0)
fnv_32_buf(fsid,..., FNV1_32_INIT)
hash32_buf(fsid,..., 0)
hash32_buf(fsid,..., HASHINIT)
- plus simply using the low order bits of fsid.val[0].
The first three behaved about equally well, with the first one being
slightly better than the others.
It has an average variation of about 4.5% about the target list length
and that is what this patch uses.
Peter Errikson also tested this hash table version and found that the
performance wasn't measurably improved by a larger hash table, so a
load factor of 10 appears adequate.
rmacklem [Fri, 31 May 2019 00:56:31 +0000 (00:56 +0000)]
Clean up silly code case.
This silly code segment has existed in the sources since it was brought
into FreeBSD 10 years ago. I honestly have no idea why this was done.
It was possible that I thought that it might have been better to not
set B_ASYNC for the "else" case, but I can't remember.
Anyhow, this patch gets rid of the if/else that does the same thing
either way, since it looks silly and upsets a static analyser.
This will have no semantic effect on the NFS client.
gjb [Thu, 30 May 2019 16:49:40 +0000 (16:49 +0000)]
Fix GCE virtual machine startup.
The ports/head branch recently switched to python3 as the default,
which breaks the sysutils/py-google-compute-engine startup scripts,
as lang/python installs lang/python3{,.x} where lang/python2{,.x}
are needed.
Set DEFAULT_VERSIONS in release/tools/gce.conf to python=2.7, and
remove the lang/python3 inclusion in VM_EXTRA_PACKAGES.
Additionally, unset DEFAULT_VERSIONS in release/tools/vmimage.subr
to prevent persistence of DEFAULT_VERSIONS=python=2.7 in subsequent
VM/cloud image builds.
Note: at present, this affects only 13-CURRENT and 12-STABLE, as
the stable/11 branch had already switched to using the 2019Q2 branch
at the start of the 11.3-RELEASE cycle, so this does not immediately
affect 11.3-BETA, hence the 1-week merge timeout. This had been
manually tested on 13-CURRENT.
Reported by: ler (privately)
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
markj [Thu, 30 May 2019 15:28:48 +0000 (15:28 +0000)]
elfcopy: Optimize for insertions at the end of the section list.
This is the common case when strip(1) is creating the output file.
The change provides a significant speedup when running on ELF files with
many sections.
PR: 234949
Reviewed by: emaste
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D20444
mav [Thu, 30 May 2019 15:07:39 +0000 (15:07 +0000)]
Pass data pointers to the driver in way in expects.
Probably due to historical reasons the driver uses In/Out arguments in
odd way. While this tool still never uses Out arguments to see that,
make the code to not trigger EINVAL in possible future uses.
kib [Thu, 30 May 2019 15:04:09 +0000 (15:04 +0000)]
Silence witness warning about duplicated mutex type.
The order is correct, it is nullfs vnode interlock -> lower vnode
interlock. vop_stdadd_writecount() is called from nullfs
VOP_ADD_WRITECOUNT() and both take interlocks.
Requested by: markj
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
mw [Thu, 30 May 2019 13:45:41 +0000 (13:45 +0000)]
Improve ENA reset handling
For easier debugging, the reset is being triggered and the reset reason is
being set only in case it is done for the first time. Such approach will
ensure that the first reset reason is not going to be overwritten and
will make it easier for debugging.
Also, add a reset trigger upon invalid Tx requested ID.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:41:39 +0000 (13:41 +0000)]
Unify new line characters in the ENA driver
Some messages were missing new line character and traces were not having
unified behavior. To fix that, each trace and printout should add new
line character at the end of each string - that should improve
readability.
Submitted by: Rafal Kozik <rk@semihalf.com>
Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:40:51 +0000 (13:40 +0000)]
Fix Tx offloads for fragmented pkt headers in ENA
If the headers of the packets are split into multiple segments of the
mbuf chain, the previous version of ena_tx_csum which was assuming,
that all segments will lay in the first mbuf, will eventually fail to
map the headers properties to meta descriptor.
That will cause Tx checksum offload to do not work and was leading to
memory corruption. It could even cause the crash of the system.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:37:15 +0000 (13:37 +0000)]
Use bitfield for storing global ENA device states
As the ENA can have multiple states turned on/off, it is more convenient
to store them in single bitfield instead of multiple boolean variables.
The bitset FreeBSD API was used for the bitfield implementation, as it
provides flexible structure together with API which also supports atomic
bitfield operations.
For better readability basic macros from API were wrapped into custom
ENA_FLAG_* macros, which are filling up common parameters for all calls.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:33:31 +0000 (13:33 +0000)]
Add additional doorbells on ENA Tx path
The new ENA HAL is introducing API, which can determine on Tx path if
the doorbell is needed.
That way, it can tell the driver, that it should call an doorbell.
The old threshold value wasn't removed, as not all HW is supporting this
feature - so it was reworked to also work with the new API.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:30:52 +0000 (13:30 +0000)]
Add support for the LLQv2 and WC in ENA
LLQ (Low Latency Queue) is the feature, that allows pushing header
directly to the device through PCI before even DMA is triggered.
It reduces latency, because device can start preparing packet before
payload is sent through DMA.
To speed up sending data through PCI, the Write Combining is enabled,
which allows hardware to buffer data before sending them on the PCI - it
allows to reduce number of PCI IO operations.
ENAv2 is using special descriptor for the negotiation of the LLQ.
Currently, only the default configuration is supported.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:24:47 +0000 (13:24 +0000)]
Change attach order to prevent crash upon failure in ENA
The if_detach was causing crash if the MSI-x configuration in the attach
failed. To prevent this issue, the ifnet is being configured at the end
of the attach function.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:22:12 +0000 (13:22 +0000)]
Check for number of MSI-x upon partial allocation in ENA
The ENA driver needs at least 2 MSI-x - one for admin queue, and one for
IO queues pair. If there were not enough resources to allocate more than
one MSI-x, the device should not be attached.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:20:42 +0000 (13:20 +0000)]
Set error value when allocation of IO irq fails in ENA
bus_alloc_resource_any() is not returning error value in case of an
error.
If the function call fails, the error value was not passed to the
ena_up() function.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:18:23 +0000 (13:18 +0000)]
Fix DMA synchronization in the ENA driver Tx and Rx paths
The DMA in FreeBSD requires explicit synchronization. ENA driver was
only doing PREREAD and PREWRITE synchronizations. Missing
bus_dmamap_sync() calls were added.
It is also required to synchronize DMA engine before unloading DMA map.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:16:56 +0000 (13:16 +0000)]
Check for missing MSI-x and Tx completions in ENA
If the first MSI-x won't be executed, then the timer service will detect
that and trigger device reset.
The checking for missing Tx completion was reworked, so it will also
check for missing interrupts. Checking number of missing Tx completions
can be performed after loop, instead of checking it every iteration.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:13:15 +0000 (13:13 +0000)]
Trigger reset in ENA if there are too many Rx descriptors
Whenever the driver will receive too many descriptors from the device,
it should trigger the device reset, as it is indicating that the device
is in invalid state.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:12:14 +0000 (13:12 +0000)]
Remove RSS support in ENA
Receive Side Scaling is optional feature that could be enabled in kernel
configuration by defining flag RSS.
Kernel uses hash to store and find protocol control block which is
stored in hash tables.
Kernel and NIC hash functions must be consistent. Otherwise case lookup
fails.
To achieve this kernel provides API to set proper hash key to NIC.
As it is not possible to change key for virtual ENA NIC, this driver
cannot support RSS function.
ENA is designed to work in virtual environments so supporting hardware
version of this card is unnecessary.
mw [Thu, 30 May 2019 13:09:53 +0000 (13:09 +0000)]
Add notification AENQ handler for ENA
Notification AENQ handler is responsible for handling requests from ENA
device. Missing Tx threshold, Tx timeout and keep alive timeout can be
set using hints from the aenq descriptor which can be delivered in the
ENA admin notification.
The queue suspending and resuming tasks are not supported by the
driver.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
mw [Thu, 30 May 2019 13:08:00 +0000 (13:08 +0000)]
Print information when ENA admin error occurs
ENA_ADMIN_FATAL_ERROR and ENA_ADMIN_WARNING aenq groups were indicated
as supported, so the unimplemented_aenq_handler() will print out error
message, whenever an error will occur within the ENA admin context.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
jchandra [Thu, 30 May 2019 01:32:00 +0000 (01:32 +0000)]
gicv3_its: do LPI init only once per CPU
The initialization required for LPIs (setting up pending tables etc.)
has to be done just once per CPU, even in the case where there are
multiple ITS blocks associated with the CPU.
Add a flag lpi_enabled in the per-cpu distributor info for this and
use it to ensure that we call its_init_cpu_lpi() just once.
This enables us to support platforms where multiple GIC ITS blocks
can generate LPIs to a CPU.
Reviewed by: andrew
Differential Revision: https://reviews.freebsd.org/D19844
jchandra [Thu, 30 May 2019 01:24:47 +0000 (01:24 +0000)]
gicv3_its: refactor LPI init into a new function
Move the per-cpu LPI intialization to a separate function. This is
in preparation for a commit that does LPI init only once for a CPU,
even when there are multiple ITS blocks associated with the CPU.
No functional changes in this commit.
Reviewed by: andrew
Differential Revision: https://reviews.freebsd.org/D19843
jchandra [Thu, 30 May 2019 01:21:08 +0000 (01:21 +0000)]
gic_v3: consolidate per-cpu redistributor information
Update 'struct gic_redists' to consolidate all per-cpu redistributor
information into a new 'struct redist_pcpu'. Provide a new interface
(GICV3_IVAR_REDIST) for the GIC driver, which can be used to retrieve
the per-cpu data.
This per-cpu redistributor struct will be later used to improve the
GIC ITS setup.
While there, remove some unused fields in gic_v3_var.h interface.
No functional changes.
Reviewed by: andrew
Differential Revision: https://reviews.freebsd.org/D19842
rpokala [Wed, 29 May 2019 23:50:31 +0000 (23:50 +0000)]
Add bits related to SANITIZE, SED, and form-factor to (struct ata_params)
Based on ATA-ACS-4, recognize several bit-fields related to the ATA SANITIZE
feature-set, Self-Encrypting Drives, and form-factor identification.
As part of this change, the name of word 48 of (struct ata_params) is being
changed. The previous name, "usedmovsd" does not appear to be related to the
previous definition of the word ("double-word IO supported"). The word was
defined that way in ATA-1 (1994), but it was marked "Reserved" (meaning
"unused, but might be used in the future") in ATA-2 (1996). It stayed that
way until ATA-8 (2008), which re-defined it as implemented in this change.
The field is not used in-tree.
erj [Wed, 29 May 2019 22:24:10 +0000 (22:24 +0000)]
iflib: provide probe wrapper for vendor drivers
From Jake:
Vendor drivers that exist out-of-tree generally should return
BUS_PROBE_VENDOR from their device probe functions. This helps ensure
that a vendor replacement driver will supersede the in-kernel driver for
a given device.
Currently, if a vendor wants to implement a driver based on iflib, it
will always report BUS_PROBE_DEFAULT.
Add a wrapper function, iflib_device_probe_vendor() which can be used in
place of iflib_device_probe(). This function will just return
BUS_PROBE_VENDOR whenever iflib_device_probe() would return
BUS_PROBE_DEFAULT.
While vendor drivers can already implement such a wrapper themselves,
providing it in the iflib.h header makes it easier for the vendor driver
to do the right thing.
allanjude [Wed, 29 May 2019 20:34:35 +0000 (20:34 +0000)]
Fix assertion in ZFS TRIM code
Due to an attempt to check two conditions at once in a macro not designed
as such, the assertion would always evaluate to true.
#define VERIFY3_IMPL(LEFT, OP, RIGHT, TYPE) do { \
const TYPE __left = (TYPE)(LEFT); \
const TYPE __right = (TYPE)(RIGHT); \
if (!(__left OP __right)) \
assfail3(#LEFT " " #OP " " #RIGHT, \
(uintmax_t)__left, #OP, (uintmax_t)__right, \
__FILE__, __LINE__); \
_NOTE(CONSTCOND) } while (0)
#define ASSERT3U(x, y, z) VERIFY3_IMPL(x, y, z, uint64_t)
Mean that we compared:
left = (type == ZIO_TYPE_FREE || psize)
OP = "<="
right = (SPA_MAXBLOCKSIZE)
If the type was not FREE, 0 is less than SPA_MAXBLOCKSIZE (16MB)
If the type is ZIO_TYPE_FREE, 1 is less than SPA_MAXBLOCKSIZE
The constraint on psize (physical size of the FREE operation) is never
checked against SPA_MAXBLOCKSIZE
Reported by: Ka Ho Ng <khng300@gmail.com>
Reviewed by: kevans
MFC after: 2 weeks
Sponsored by: Klara Systems
kib [Wed, 29 May 2019 14:05:27 +0000 (14:05 +0000)]
Do not go into sleep in sleepq_catch_signals() when SIGSTOP from
PT_ATTACH was consumed.
In particular, do not clear TDP_FSTP in ptracestop() if td_wchan is
non-NULL. Leave it to sleepq_catch_signal() to clear and convert zero
return code to EINTR.
Otherwise, per submitter report, if the PT_ATTACH SIGSTOP was
delivered right after the thread was added to the sleepqueue but not
yet really sleep, and cursig() caused debugger attach, the thread
sleeps instead of returning to the userspace boundary with EINTR.
avg [Wed, 29 May 2019 09:08:20 +0000 (09:08 +0000)]
revert r273728 and parts of r306589, iicbus no-stop by default feature
Since drm2 removal, there has not been any consumer of the feature in the
tree. I am also unaware of any out-of-tree consumer.
More importantly, the feature has been broken from the very start, both
before and after r306589, because the ivar was set on a device that does
not support it and it was read from another device that also does not
support it.
A bus-wide no-stop flag cannot be implemented as an ivar as iicbus
attaches as a child of various drivers. Implementing the ivar in each
and every I2C driver is just impractical.
If we ever want to implement this feature properly, then probably the
easiest way to do it would be via a flag in the softc of iicbus.
In fact, we might have to do that in the stable branches if we want to
fix the code for them.
Reported by: ian (long time ago)
MFC after: 1 month (maybe)
X-MFC-note: cannot just merge the change, must keep drm2 happy
jhibbits [Wed, 29 May 2019 02:02:56 +0000 (02:02 +0000)]
Add missing powerpc64 relocation support to libdwarf
Summary:
Due to missing relocation support in libdwarf for powerpc64, handling of dwarf
info on unlinked objects was bogus.
Examining raw dwarf data on objects compiled on ppc64 with a modern compiler
(in-tree gcc tends to hide the issue, since it only rarely generates relocations
in .debug_info and uses DW_FORM_str instead of DW_FORM_strp for everything), you
will find that the dwarf data appears corrupt, with repeated references to the
compiler version where things like types and function names should appear.
This happens because the 0 offset of .debug_str contains the compiler version,
and without applying the relocations, *all* indirect strings in .dwarf_info will
end up pointing to it.
This corruption then propogates to the CTF data, as ctfconvert relies on
libdwarf to read the dwarf info, for every compiled object (when building a
kernel.)
However, if you examine the dwarf data on a compiled executable, it will appear
correct, because during final link the relocations get applied and baked in by
the linker.
kevans [Wed, 29 May 2019 01:08:30 +0000 (01:08 +0000)]
if_bridge(4): Complete bpf auditing of local traffic over the bridge
There were two remaining "gaps" in auditing local bridge traffic with
bpf(4):
Locally originated outbound traffic from a member interface is invisible to
the bridge's bpf(4) interface. Inbound traffic locally destined to a member
interface is invisible to the member's bpf(4) interface -- this traffic has
no chance after bridge_input to otherwise pass it over, and it wasn't
originally received on this interface.
I call these "gaps" because they don't affect conventional bridge setups.
Alas, being able to establish an audit trail of all locally destined traffic
for setups that can function like this is useful in some scenarios.
johalun [Tue, 28 May 2019 20:54:59 +0000 (20:54 +0000)]
pseudofs: Ignore unsupported commands in vop_setattr.
Users of pseudofs (e.g. lindebugfs), should be able to receive
input from command line via commands like "echo 1 > /path/to/file".
Currently this fails because sh tries to truncate the file first and
vop_setattr returns not supported error for this. This patch simply
ignores the error and returns 0 instead.
mav [Tue, 28 May 2019 18:32:04 +0000 (18:32 +0000)]
Fix array out of bound panic introduced in r306219.
As I see, different NICs in different configurations may have different
numbers of TX and RX queues. The code was assuming 1:1 mapping between
event queues (interrupts) and TX/RX queues. Since number of interrupts
is set to maximum of TX and RX queues, when those two are different, the
system is doomed.
I have no documentation or deep knowledge about this hardware, so this
change is based on general observations and code reading. If some of my
guesses are wrong, please do better. I just confirmed HP NC550SFP NICs
are working now.
kevans [Tue, 28 May 2019 16:12:16 +0000 (16:12 +0000)]
bectl(8): Address Coverity complaints
CID 1400451: case 0 is missing a break/return and falling through to the
default case. waitpid(0, ...) makes little sense in the child, we likely
wanted to terminate immediately.
CID 1400453: size argument uses sizeof(char **) instead of sizeof(char *)
and is assigned to a char **; sizeof's match but "this isn't a portable
assumption".
dougm [Tue, 28 May 2019 15:47:00 +0000 (15:47 +0000)]
Implement the ffs and fls functions, and their longer counterparts, in
cpufunc, in terms of __builtin_ffs and the like, for arm32 v6 and v7
architectures, and use those, rather than the simple libkern
implementations, in building arm32 kernels.
Reviewed by: manu
Approved by: kib, markj (mentors)
Tested by: iz-rpi03_hs-karlsruhe.de, mikael.urankar_gmail.com, ian
Differential Revision: https://reviews.freebsd.org/D20412
ae [Tue, 28 May 2019 11:45:00 +0000 (11:45 +0000)]
Rework r348303 to reduce the time of holding global BPF lock.
It appeared that using NET_EPOCH_WAIT() while holding global BPF lock
can lead to another panic:
spin lock 0xfffff800183c9840 (turnstile lock) held by 0xfffff80018e2c5a0 (tid 100325) too long
panic: spin lock held too long
...
#0 sched_switch (td=0xfffff80018e2c5a0, newtd=0xfffff8000389e000, flags=<optimized out>) at /usr/src/sys/kern/sched_ule.c:2133
#1 0xffffffff80bf9912 in mi_switch (flags=256, newtd=0x0) at /usr/src/sys/kern/kern_synch.c:439
#2 0xffffffff80c21db7 in sched_bind (td=<optimized out>, cpu=<optimized out>) at /usr/src/sys/kern/sched_ule.c:2704
#3 0xffffffff80c34c33 in epoch_block_handler_preempt (global=<optimized out>, cr=0xfffffe00005a1a00, arg=<optimized out>)
at /usr/src/sys/kern/subr_epoch.c:394
#4 0xffffffff803c741b in epoch_block (global=<optimized out>, cr=<optimized out>, cb=<optimized out>, ct=<optimized out>)
at /usr/src/sys/contrib/ck/src/ck_epoch.c:416
#5 ck_epoch_synchronize_wait (global=0xfffff8000380cd80, cb=<optimized out>, ct=<optimized out>) at /usr/src/sys/contrib/ck/src/ck_epoch.c:465
#6 0xffffffff80c3475e in epoch_wait_preempt (epoch=0xfffff8000380cd80) at /usr/src/sys/kern/subr_epoch.c:513
#7 0xffffffff80ce970b in bpf_detachd_locked (d=0xfffff801d309cc00, detached_ifp=<optimized out>) at /usr/src/sys/net/bpf.c:856
#8 0xffffffff80ced166 in bpf_detachd (d=<optimized out>) at /usr/src/sys/net/bpf.c:836
#9 bpf_dtor (data=0xfffff801d309cc00) at /usr/src/sys/net/bpf.c:914
To fix this add the check to the catchpacket() that BPF descriptor was
not detached just before we acquired BPFD_LOCK().
andrew [Tue, 28 May 2019 10:55:59 +0000 (10:55 +0000)]
The alignment is passed into contigmalloc_domainset in the 7th argument.
KUBSAN was complaining the pointer contigmalloc_domainset returned was
misaligned. Fix this by using the correct argument to find the alignment
in the function signature.