Marcin Wojtas [Tue, 26 May 2020 15:57:02 +0000 (15:57 +0000)]
Add sysctl node for ENA IO queues number adjustment
By default, in ena_attach() the driver attempts to acquire
ena_adapter::max_num_io_queues MSI-X vectors for the purpose of IO
queues, however this is not guaranteed. The number of vectors acquired
depends also on system resources availability.
Regardless of that, enable the number of effectively used IO queues to
be further limited through the sysctl node.
Example: Assumming that there are 8 IO queues configured by default, the
command
$ sysctl dev.ena.0.io_queues_nb=4
will reduce the number of available IO queues to 4. Similarly, the value
can be also increased up to maximum supported value. A value higher than
maximum supported number of IO queues is ignored. Zero is ignored too.
Submitted by: Maciej Bielski <mba@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
Marcin Wojtas [Tue, 26 May 2020 15:54:32 +0000 (15:54 +0000)]
Fix assumptions about number of IO queues in the ENA
Make the ena_adapter::num_io_queues a number of effectively used IO
queues. While the ena_adapter::max_num_io_queues is an upper-bound
specified by the HW, the ena_adapter::num_io_queues may be lower than
that, depending on runtime system resources availability.
On reset, there are called ena_destroy_device() and then
ena_restore_device(). The latter calls, in turn, ena_enable_msix(),
which will attempt to re-acquire ena_adapter::max_num_io_queues of
MSIX vectors again.
Thus, the value of ena_adapter::num_io_queues may be different before
and after reset. For this reason, free the IO rings structures (drbr,
counters) in ena_destroy_device() and allocate again in
ena_restore_device().
Submitted by: Maciej Bielski <mba@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
Marcin Wojtas [Tue, 26 May 2020 15:48:06 +0000 (15:48 +0000)]
Rework ENA Rx queue size configuration
This patch reworks how the Rx queue size is being reconfigured and how
the information from the device is being processed.
Reconfiguration of the queues and reset of the device in order to make
the changes alive isn't the best approach. It can be done synchronously
and it will let to pass information if the reconfiguration was
successful to the user. It now is done in the ena_update_queue_size()
function.
To avoid reallocation of the ring buffer, statistic counters and the
reinitialization of the mutexes when only new size has to be assigned,
the io queues initialization function has been split into 2 stages:
basic, which is just copying appropriate fields and the advanced, which
allocates and inits more advanced structures for the IO rings.
Moreover, now the max allowed Rx and Tx ring size is being kept
statically in the adapter and the size of the variables holding those
values has been changed to uint32_t everywhere.
Information about IO queues size is now being logged in the up routine
instead of the attach.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
Marcin Wojtas [Tue, 26 May 2020 15:45:54 +0000 (15:45 +0000)]
Mark the ENA driver as epoch ready
Recent changes to the epoch requires driver to notify that they knows
epoch in order to prevent input packet function to enter epoch each
time the packet is received.
ENA is using NET_TASK for handling Rx, so it's entering epoch
automatically whenever this task is being executed.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
Marcin Wojtas [Tue, 26 May 2020 15:41:53 +0000 (15:41 +0000)]
Expose argument names for non static ENA driver functions
As functions which are declared in the header files are intended to be
the interface and are going to be used by other files, it's better to
include argument names in the definition, so the caller won't have to
check the .c file in order to check their meaning and order.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
Marcin Wojtas [Tue, 26 May 2020 15:39:41 +0000 (15:39 +0000)]
Use single global lock in the ENA driver
Currently, the driver had 2 global locks - one was sx lock used for
up/down synchronization and the second one was mutex, which was used
for link configuration and timer service callout.
It is better to have single lock for that. We cannot use mutex, as it
can sleep and cause witness errors in up/down configuration, so sx lock
seems to be the only choice.
Callout cannot use sx lock, but the timer service is MP safe, so we just
need to avoid race between ena_down() and ena_detach(). It can be
avoided by acquiring sx lock.
Simple macros were added that are encapsulating implementation of the
lock and makes the code cleaner.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
Marcin Wojtas [Tue, 26 May 2020 15:37:55 +0000 (15:37 +0000)]
Add trigger reset function in the ENA driver
As the reset triggering is no longer a simple macro that was just
setting appropriate flag, the new function for triggering reset was
added. It improves code readability a lot, as we are avoiding additional
indentation.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
Marcin Wojtas [Tue, 26 May 2020 15:33:43 +0000 (15:33 +0000)]
Remove unused argument from static function in ena.c
The function ena_enable_msix_and_set_admin_interrupts takes two
arguments while the second is not used and so can be spared. This is a
static function, only ena.c is affected.
Submitted by: Maciej Bielski <mba@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
Marcin Wojtas [Tue, 26 May 2020 15:31:28 +0000 (15:31 +0000)]
Enable Tx drops reporting in the ENA driver
Tx drops statistics are fetched from HW every ena_keepalive_wd() call
and are observable using one of the commands:
* sysctl dev.ena.0.hw_stats.tx_drops
* netstat -I ena0 -d
Submitted by: Maciej Bielski <mba@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
Marcin Wojtas [Tue, 26 May 2020 15:29:19 +0000 (15:29 +0000)]
Adjust ENA driver to the new HAL
* Removed adaptive interrupt moderation (not suported on FreeBSD).
* Use ena_com_free_q_entries instead of ena_com_free_desc.
* Don't use ENA_MEM_FREE outside of the ena_com.
* Don't use barriers before calling doorbells as it's already done in
the HAL.
* Add function that generates random RSS key, common for all driver's
interfaces.
* Change admin stats sysctls to U64.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
Marcin Wojtas [Tue, 26 May 2020 15:04:49 +0000 (15:04 +0000)]
Upgrade ENA HAL to the version from 20.04.2020
This version of the HAL supports newest generation ENA HW, random
RSS generation upon device initialization and also includes bug fixes
in the platform file, like fix for IO write/read macros regarding using
barriers.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
Marcin Wojtas [Tue, 26 May 2020 14:16:26 +0000 (14:16 +0000)]
Fix AES-CTR compatibility issue in ipsec
r361390 decreased blocksize of AES-CTR from 16 to 1.
Because of that ESP payload is no longer aligned to 16 bytes
before being encrypted and sent.
This is a good change since RFC3686 specifies that the last block
doesn't need to be aligned.
Since FreeBSD before r361390 couldn't decrypt partial blocks encrypted
with AES-CTR we need to enforce 16 byte alignment in order to preserve
compatibility.
Add a sysctl(on by default) to control it.
Marcin Wojtas [Tue, 26 May 2020 14:10:53 +0000 (14:10 +0000)]
Restore XHCI operation on Armada 38x
r347343 split generic xhci driver into three files.
Include generic_xhci_fdt.c when building kernel for Armada SoCs.
This brings back XHCI support on these platforms and also
others, which use GENERIC config.
Roger Pau Monné [Tue, 26 May 2020 10:24:06 +0000 (10:24 +0000)]
xen-locore: fix size in GDT descriptor
There was an off-by-one in the GDT descriptor size field used by the
early Xen boot code. The GDT descriptor size should be the size of the
GDT minus one. No functional change expected as a result of this
change.
Justin Hibbits [Tue, 26 May 2020 03:58:19 +0000 (03:58 +0000)]
powerpc/booke pmap: Fix iteration for 64-bit kernel page table creation
Kernel page tables actually start at index 4096, given kernel base address
of 0xc008000000000000, not index 0, which would yield 0xc000000000000000.
Fix this by indexing at the real base, instead of the assumed base.
Chuck Silvers [Mon, 25 May 2020 23:47:31 +0000 (23:47 +0000)]
This commit enables a UFS filesystem to do a forcible unmount when
the underlying media fails or becomes inaccessible. For example
when a USB flash memory card hosting a UFS filesystem is unplugged.
The strategy for handling disk I/O errors when soft updates are
enabled is to stop writing to the disk of the affected file system
but continue to accept I/O requests and report that all future
writes by the file system to that disk actually succeed. Then
initiate an asynchronous forced unmount of the affected file system.
There are two cases for disk I/O errors:
- ENXIO, which means that this disk is gone and the lower layers
of the storage stack already guarantee that no future I/O to
this disk will succeed.
- EIO (or most other errors), which means that this particular
I/O request has failed but subsequent I/O requests to this
disk might still succeed.
For ENXIO, we can just clear the error and continue, because we
know that the file system cannot affect the on-disk state after we
see this error. For EIO or other errors, we arrange for the geom_vfs
layer to reject all future I/O requests with ENXIO just like is
done when the geom_vfs is orphaned. In both cases, the file system
code can just clear the error and proceed with the forcible unmount.
This new treatment of I/O errors is needed for writes of any buffer
that is involved in a dependency. Most dependencies are described
by a structure attached to the buffer's b_dep field. But some are
created and processed as a result of the completion of the dependencies
attached to the buffer.
Clearing of some dependencies require a read. For example if there
is a dependency that requires an inode to be written, the disk block
containing that inode must be read, the updated inode copied into
place in that buffer, and the buffer then written back to disk.
Often the needed buffer is already in memory and can be used. But
if it needs to be read from the disk, the read will fail, so we
fabricate a buffer full of zeroes and pretend that the read succeeded.
This zero'ed buffer can be updated and written back to disk.
The only case where a buffer full of zeros causes the code to do
the wrong thing is when reading an inode buffer containing an inode
that still has an inode dependency in memory that will reinitialize
the effective link count (i_effnlink) based on the actual link count
(i_nlink) that we read. To handle this case we now store the i_nlink
value that we wrote in the inode dependency so that it can be
restored into the zero'ed buffer thus keeping the tracking of the
inode link count consistent.
Because applications depend on knowing when an attempt to write
their data to stable storage has failed, the fsync(2) and msync(2)
system calls need to return errors if data fails to be written to
stable storage. So these operations return ENXIO for every call
made on files in a file system where we have otherwise been ignoring
I/O errors.
John Baldwin [Mon, 25 May 2020 23:04:18 +0000 (23:04 +0000)]
Expand coverage of different buffer sizes.
- When -z is used, include small buffers from 1 to 32 bytes to test
stream ciphers. Note that while AES-XTS claims to support a block
size of 1 in OpenSSL, it does require a minimum of 1 block of cipher
text as it is not a stream cipher but depends on CTS to pad out the
final partial block.
- Permit multiple AAD sizes to be set via multiple -A options, or via
-z. When -z is set, use small buffers from 0 to 32 bytes followed
by powers of 2 up to 256. When multiple sizes are specified, the
ETA and AEAD algorithms perform the full matrix of AAD sizes by
payload sizes.
- Only warn on unchanged ciphertext instead of erroring. The
currently generated plaintext and key for a couple of AES-CTR tests
with a buffer size of 1 results in ciphertext that matches the
plaintext.
Reviewed by: cem
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D25006
Adrian Chadd [Mon, 25 May 2020 22:31:45 +0000 (22:31 +0000)]
[ath] [ath_hal] Propagate the HAL_RESET_TYPE through to the chip reset; set it during ath_reset()
Although I added the reset type field to ath_hal_reset() years ago,
I never finished adding it both throughout the HALs and in if_ath.c.
This will eventually deprecate the ath_hal force_full_reset option
because it can be requested at the driver layer.
So:
* Teach ar5416ChipReset() and ar9300_chip_reset() about the HAL type
* Use it in ar5416Reset() and ar9300_reset() when doing a full chip reset
* Extend ath_reset() to include the HAL_RESET_TYPE parameter added in the above functions
* Use HAL_RESET_NORMAL in most calls to ath_reset()
* .. but use HAL_RESET_BBPANIC for the BB panics, and HAL_RESET_FORCE_COLD during fatal, beacon miss and other hardware related hangs.
This should be a glorified no-op outside of actual hardware issues.
I've tested things with ath_hal force_full_reset set to 1 for years now,
so I know that feature and a full reset works (albeit much slower than
a warm reset!) and it does unwedge hardware.
The eventual aim is to use this for all the places where the driver
detects a potential hang as well as if long calibration - ie, noise floor
calibration - fails to complete. That's one of the big hardware related
things that causes station mode operation to hang without easy recovery.
John Baldwin [Mon, 25 May 2020 22:30:44 +0000 (22:30 +0000)]
Support separate output buffers for aesni(4).
The backend routines aesni(4) call for specific encryption modes all
expect virtually contiguous input/output buffers. If the existing
output buffer is virtually contiguous, always write to the output
buffer directly from the mode-specific routines. If the output buffer
is not contiguous, then a temporary buffer is allocated whose output
is then copied to the output buffer. If the input buffer is not
contiguous, then the existing buffer used to hold the input is also
used to hold temporary output.
John Baldwin [Mon, 25 May 2020 22:12:04 +0000 (22:12 +0000)]
Add support for optional separate output buffers to in-kernel crypto.
Some crypto consumers such as GELI and KTLS for file-backed sendfile
need to store their output in a separate buffer from the input.
Currently these consumers copy the contents of the input buffer into
the output buffer and queue an in-place crypto operation on the output
buffer. Using a separate output buffer avoids this copy.
- Create a new 'struct crypto_buffer' describing a crypto buffer
containing a type and type-specific fields. crp_ilen is gone,
instead buffers that use a flat kernel buffer have a cb_buf_len
field for their length. The length of other buffer types is
inferred from the backing store (e.g. uio_resid for a uio).
Requests now have two such structures: crp_buf for the input buffer,
and crp_obuf for the output buffer.
- Consumers now use helper functions (crypto_use_*,
e.g. crypto_use_mbuf()) to configure the input buffer. If an output
buffer is not configured, the request still modifies the input
buffer in-place. A consumer uses a second set of helper functions
(crypto_use_output_*) to configure an output buffer.
- Consumers must request support for separate output buffers when
creating a crypto session via the CSP_F_SEPARATE_OUTPUT flag and are
only permitted to queue a request with a separate output buffer on
sessions with this flag set. Existing drivers already reject
sessions with unknown flags, so this permits drivers to be modified
to support this extension without requiring all drivers to change.
- Several data-related functions now have matching versions that
operate on an explicit buffer (e.g. crypto_apply_buf,
crypto_contiguous_subsegment_buf, bus_dma_load_crp_buf).
- Most of the existing data-related functions operate on the input
buffer. However crypto_copyback always writes to the output buffer
if a request uses a separate output buffer.
- For the regions in input/output buffers, the following conventions
are followed:
- AAD and IV are always present in input only and their
fields are offsets into the input buffer.
- payload is always present in both buffers. If a request uses a
separate output buffer, it must set a new crp_payload_start_output
field to the offset of the payload in the output buffer.
- digest is in the input buffer for verify operations, and in the
output buffer for compute operations. crp_digest_start is relative
to the appropriate buffer.
- Add a crypto buffer cursor abstraction. This is a more general form
of some bits in the cryptosoft driver that tried to always use uio's.
However, compared to the original code, this avoids rewalking the uio
iovec array for requests with multiple vectors. It also avoids
allocate an iovec array for mbufs and populating it by instead walking
the mbuf chain directly.
- Update the cryptosoft(4) driver to support separate output buffers
making use of the cursor abstraction.
Conrad Meyer [Mon, 25 May 2020 16:40:48 +0000 (16:40 +0000)]
copystr(9): Move to deprecate (attempt #2)
This reapplies logical r360944 and r360946 (reverting r360955), with fixed
copystr() stand-in replacement macro. Eventually the goal is to convert
consumers and kill the macro, but for a first step it helps if the macro is
correct.
Prior commit message:
Unlike the other copy*() functions, it does not serve to copy from one
address space to another or protect against potential faults. It's just
an older incarnation of the now-more-common strlcpy().
Add a coccinelle script to tools/ which can be used to mechanically
convert existing instances where replacement with strlcpy is trivial.
In the two cases which matched, fuse_vfsops.c and union_vfsops.c, the
code was further refactored manually to simplify.
Replace the declaration of copystr() in systm.h with a small macro
wrapper around strlcpy (with correction from brooks@ -- thanks).
Remove N redundant MI implementations of copystr. For MIPS, this
entailed inlining the assembler copystr into the only consumer,
copyinstr, and making the latter a leaf function.
Marcin Wojtas [Mon, 25 May 2020 15:40:02 +0000 (15:40 +0000)]
Introduce support for Epson RX-8803 RTC.
This patch introduces support for Epson RX-8803 RTC controller accessible
over I2C bus. It has a resolution of 1 sec.
Support for interrupt based alarm was not implemented.
Submitted by: Kornel Duleba <mindal@semihalf.com>
Reviewed by: manu
Obtained from: Semihalf
Sponsored by: Alstom Group
Differential Revision: https://reviews.freebsd.org/D24364
Marcin Wojtas [Mon, 25 May 2020 15:31:43 +0000 (15:31 +0000)]
Add TCA6416 GPIO expander support.
Add basic TCA6416 GPIO expander support over I2C bus. The driver handles
enabling and disabling pins, setting pin mode to IN and OUT and
toggling the pins. External interrupts are not supported.
Marcin Wojtas [Mon, 25 May 2020 15:21:38 +0000 (15:21 +0000)]
Introduce VF610 I2C controller support.
NXP LS1046A contains I2C controller compatible with Vybrid VF610.
Existing Vybrid MVF600 driver can be used to support it. For that purpose
declare driver as ofw_iicbus and add methods associated with ofw_iicbus.
For VF610 add dynamic clock prescaler calculation using clock information
from clock driver and clock frequency requested in device tree.
On the occasion add detach function and add additional error handling
in i2c_attach function.
Submitted by: Dawid Gorecki <dgr@semihalf.com>
Reviewed by: manu
Obtained from: Semihalf
Sponsored by: Alstom Group
Differential Revision: https://reviews.freebsd.org/D24361
Marcin Wojtas [Mon, 25 May 2020 14:55:37 +0000 (14:55 +0000)]
Add GPIO support for QorIQ boards.
This patch adds a GPIO controller support targeted for NXP LS1046A
SoC. The driver implements the following features:
* setting direction of each pin (IN or OUT)
* setting the mode of output pins (PUSHPULL or OPENDRAIN)
* setting the state of each output pin (1 or 0)
* reading the state of each input pin (1 or 0)
Marcin Wojtas [Mon, 25 May 2020 14:45:18 +0000 (14:45 +0000)]
Add LS1046A clockgen driver.
Driver provides probe and attach functions for LS1046A clockgen and passes
configuration information to QorIQ clockgen class. It may be used as
a reference implementation for different QorIQ clockgen devices.
Submitted by: Dawid Gorecki <dgr@semihalf.com>
Reviewed by: mmel, manu
Obtained from: Semihalf
Sponsored by: Alstom Group
Differential Revision: https://reviews.freebsd.org/D24352
Marcin Wojtas [Mon, 25 May 2020 14:31:32 +0000 (14:31 +0000)]
Add QorIQ platform clockgen driver.
This patch adds classes and functions that can be used with various NXP
QorIQ Layerscape SoCs.
As for the clock topology - there is single platform PLL, which supplies
clocks for the peripheral bus and additional PLLs for CPU cores. There
can be multiple core PLLs (For example - LS1046A has two PLLs - CGAPLL1
and CGAPLL2). Each PLL has fixed dividers on output. The core PLLs
are not accessible from dts.
This is a preparation patch for NXP LS1046A SoC support.
Emmanuel Vadot [Mon, 25 May 2020 12:46:05 +0000 (12:46 +0000)]
linuxkpi: Fix mod_timer and del_timer_sync
mod_timer is supposed to return 1 if the modified timer was pending, which
is exactly what callout_reset does so return the value after checking
that it's a correct one in case the api change.
del_timer_sync returns int so add a function and handle that.
Emmanuel Vadot [Mon, 25 May 2020 12:42:55 +0000 (12:42 +0000)]
linuxkpi: Add __same_type and __must_be_array macros
The same_type macro simply wraps around builtin_types_compatible_p which
exist for both GCC and CLANG, which returns 1 if both types are the same.
The __must_be_array macros returns 1 if the argument is an array.
This is needed for DRM v5.3
Sponsored-by: The FreeBSD Foundation
Reviewed by: hselasky
Differential Revision: https://reviews.freebsd.org/D24953
Correctly set the initial vector for TLS v1.3 for mlx5en(4).
For TLS v1.3 the 12 bytes of the initial vector, IV, should just be copied
as-is from the kernel to the gcm_iv field, which hold the first 4 bytes,
and the remaining 8 bytes go to the subsequent implicit_iv field.
There is no need to consider the byte order on the 12 bytes of IV like
initially done.
Murray Stokely [Mon, 25 May 2020 07:18:47 +0000 (07:18 +0000)]
Add example usage for formatting a floppy disk. Adding a more self
contained example here in the fdformat man page will allow us to
modernize and streamline the FreeBSD Handbook by cutting out some of
this legacy material.
While here, address some other minor grammatical nits in this man page.
Peter Grehan [Mon, 25 May 2020 06:25:31 +0000 (06:25 +0000)]
Fix pci-passthru MSI issues with OpenBSD guests
- Return 2 x 16-bit registers in the correct byte order
for a 4-byte read that spans the CMD/STATUS register.
This reversal was hiding the capabilities-list, which prevented
the MSI capability from being found for XHCI passthru.
- Reorganize MSI/MSI-x config writes so that a 4-byte write at the
capability offset would have the read-only portion skipped.
This prevented MSI interrupts from being enabled.
Reported and extensively tested by Anatoli (me at anatoli dot ws)
PR: 245392
Reported by: Anatoli (me at anatoli dot ws)
Reviewed by: jhb (bhyve)
Approved by: jhb, bz (mentor)
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D24951
Ed Maste [Sun, 24 May 2020 18:25:49 +0000 (18:25 +0000)]
Make i386 memstick images bootable.
This reverts the i386 part of r342283, "Rework UEFI ESP generation", and
the followup commit in r342690.
r342283 added an ESP to the i386 memstick image, and as a side effect
made the ESP the active partition, not the bootcode-containing UFS
partition. As a result the i386 memstick images would not boot in
either UEFI or legacy mode - UEFI failed because we do not support i386
UEFI booting, and legacy mode failed because the partition with legacy
bootcode was not active.
The bootcode-containing UFS partition is again the only, and active,
partition.
PR: 246494
Reported by: Jorge Maidana
Differential Revision: The FreeBSD Foundation
Andriy Gapon [Sun, 24 May 2020 14:54:21 +0000 (14:54 +0000)]
libprocstat: try to fix fallout from r361363
The revision caused libprocstat to have two undefined symbols:
- __start_set_pcpu
- __stop_set_pcpu
probably because of __GLOBL() used in sys/pcpu.h under _KERNEL.
The symbols are not accessed by anything and the linker in base does not
complain about them, but some ports are failing to build.
Hack around the problem by providing definitions for those symbols.
Probably there is a better solution, but I could not think of it yet.
* Improved decompress performance on amd64 and arm (5-10%
and 15-50%, respectively).
* '--patch-from' zstd(1) CLI option, which provides something like a very fast
version of bspatch(1) with slightly worse compression. See release notes.
In this update, I dropped the 3-year old -O0 workaround for an LLVM ARM bug;
the bug was fixed in LLVM SVN in 2017, but we didn't remove this workaround
from our tree until now.
Move <add|del|change>_route() functions to route_ctl.c in preparation of
multipath control plane changed described in D24141.
Currently route.c contains core routing init/teardown functions, route table
manipulation functions and various helper functions, resulting in >2KLOC
file in total. This change moves most of the route table manipulation parts
to a dedicated file, simplifying planned multipath changes and making
route.c more manageable.
Use epoch(9) for rtentries to simplify control plane operations.
Currently the only reason of refcounting rtentries is the need to report
the rtable operation details immediately after the execution.
Delaying rtentry reclamation allows to stop refcounting and simplify the code.
Additionally, this change allows to reimplement rib_lookup_info(), which
is used by some of the customers to get the matching prefix along
with nexthops, in more efficient way.
The change keeps per-vnet rtzone uma zone. It adds nh_vnet field to
nhop_priv to be able to reliably set curvnet even during vnet teardown.
Rest of the reference counting code will be removed in the D24867 .
Mitchell Horne [Fri, 22 May 2020 18:54:56 +0000 (18:54 +0000)]
Simplify the RISC-V kernel linker invocation
Remove our custom SYSTEM_LD definition. This generates program headers
that are more consistent with other architectures, and more importantly,
are in line with what loader(8) expects when loading a kernel.
As noted in https://reviews.freebsd.org/D22920, there is no apparent
reason why the kernel would need a writable text segment, so removal of
the -N flag isn't likely to cause issue.
Alan Somers [Fri, 22 May 2020 18:11:17 +0000 (18:11 +0000)]
Fix issues with FUSE_ACCESS when default_permissions is disabled
This patch fixes two issues relating to FUSE_ACCESS when the
default_permissions mount option is disabled:
* VOP_ACCESS() calls with VADMIN set should never be sent to a fuse server
in the form of FUSE_ACCESS operations. The FUSE protocol has no equivalent
of VADMIN, so we must evaluate such things kernel-side, regardless of the
default_permissions setting.
* The FUSE protocol only requires FUSE_ACCESS to be sent for two purposes:
for the access(2) syscall and to check directory permissions for
searchability during lookup. FreeBSD sends it much more frequently, due to
differences between our VFS and Linux's, for which FUSE was designed. But
this patch does eliminate several cases not required by the FUSE protocol:
* for any FUSE_*XATTR operation
* when creating a new file
* when deleting a file
* when setting timestamps, such as by utimensat(2).
* Additionally, when default_permissions is disabled, this patch removes one
FUSE_GETATTR operation when deleting a file.
PR: 245689
Reported by: MooseFS FreeBSD Team <freebsd@moosefs.pro>
Reviewed by: cem
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D24777
Alexander Motin [Fri, 22 May 2020 18:10:46 +0000 (18:10 +0000)]
Do not try to fill socket send buffer to the last byte.
Setting so_snd.sb_lowat to at least 1/8 of the socket buffer size allows
send thread more actively use PDUs coalescing, that dramatically reduces
TCP lock congestion and number of context switches, when the socket is
full and PDUs are small.
Alan Somers [Fri, 22 May 2020 18:03:14 +0000 (18:03 +0000)]
Disable nullfs cacheing on top of fusefs
Nullfs cacheing can keep a large number of vnodes active. That results in
more active FUSE file handles, causing some FUSE servers to use extra
resources. Disable nullfs cacheing for fusefs, just like we already do for
NFSv4.
John Baldwin [Fri, 22 May 2020 16:29:09 +0000 (16:29 +0000)]
Improve support for stream ciphers in the software encryption interface.
Add a 'native_blocksize' member to 'struct enc_xform' that ciphers can
use if they support a partial final block. This is particular useful
for stream ciphers, but can also apply to other ciphers. cryptosoft
will only pass in native blocks to the encrypt and decrypt hooks. For
the final partial block, 'struct enc_xform' now has new
encrypt_last/decrypt_last hooks which accept the length of the final
block. The multi_block methods are also retired.
Mark AES-ICM (AES-CTR) as a stream cipher. This has some interesting
effects on IPsec in that FreeBSD can now properly receive all packets
sent by Linux when using AES-CTR, but FreeBSD can no longer
interoperate with OpenBSD and older verisons of FreeBSD which assume
AES-CTR packets have a payload padded to a 16-byte boundary. Kornel
has offered to work on a patch to add a compatiblity sysctl to enforce
additional padding for AES-CTR in esp_output to permit compatibility
with OpenBSD and older versions of FreeBSD.
AES-XTS continues to use a block size of a single AES block length.
It is possible to adjust it to support partial final blocks by
implementing cipher text stealing via encrypt_last/decrypt_last hooks,
but I have not done so.
Reviewed by: cem (earlier version)
Tested by: Kornel Dulęba <mindal@semihalf.com> (AES-CTR with IPsec)
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D24906
Andriy Gapon [Fri, 22 May 2020 11:25:45 +0000 (11:25 +0000)]
net80211: post RTM_IFINFO notification after toggling IFF_DRV_RUNNING
This is useful when a wireless driver is stopped or started in response
to events like an RF Kill button press. Applications like
wpa_supplicant depend on such events to have a correct view of interface
state.
Andriy Gapon [Fri, 22 May 2020 11:20:23 +0000 (11:20 +0000)]
libprocstat: fix ZFS support
First of all, znode_phys_t hasn't been used for storing file attributes
for a long time now. Modern ZFS versions use a System Attribute table
with a flexible layout. But more importantly all the required
information is available in znode_t itself.
It's not easy to include zfs_znode.h in userland without breaking code
because the most interesting parts of the header are kernel-only. And
hardcoding field offsets is too fragile. So, I created a new
compilation unit that includes zfs_znode.h using some mild kludges to
get it and its dependencies to compile in userland. The compilation
unit exports interesting field offsets and does not have any other code.
DCTCP: update alpha only once after loss recovery.
In mixed ECN marking and loss scenarios it was found, that
the alpha value of DCTCP is updated two times. The second
update happens with freshly initialized counters indicating
to ECN loss. Overall this leads to alpha not adjusting as
quickly as expected to ECN markings, and therefore lead to
excessive loss.