]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
5 years agosfxge(4): support RxDP scatter disabled truncate counter
Andrew Rybchenko [Thu, 29 Nov 2018 06:46:44 +0000 (06:46 +0000)]
sfxge(4): support RxDP scatter disabled truncate counter

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18242

5 years agosfxge(4): generate Medford2 RxDP stats
Andrew Rybchenko [Thu, 29 Nov 2018 06:46:33 +0000 (06:46 +0000)]
sfxge(4): generate Medford2 RxDP stats

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18241

5 years agosfxge(4): get max supported value for action MARK
Andrew Rybchenko [Thu, 29 Nov 2018 06:46:21 +0000 (06:46 +0000)]
sfxge(4): get max supported value for action MARK

The mark value for MATCH_ACTION_MARK has a maximum value.
Requesting a value larger than the maximum will cause the
filter insertion to fail with EINVAL. This patch allows the
driver to check the value at the filter validation.

Submitted by:   Roman Zhukov <roman.zhukov at oktetlabs.ru>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18240

5 years agosfxge(4): support MARK and FLAG actions in filters
Andrew Rybchenko [Thu, 29 Nov 2018 06:46:10 +0000 (06:46 +0000)]
sfxge(4): support MARK and FLAG actions in filters

This patch adds support for DPDK rte_flow "MARK" and "FLAG" filter
actions to filters on EF10 family NICs.

Submitted by:   Roman Zhukov <roman.zhukov at oktetlabs.ru>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18239

5 years agosfxge(4): get actions MARK and FLAG support
Andrew Rybchenko [Thu, 29 Nov 2018 06:46:01 +0000 (06:46 +0000)]
sfxge(4): get actions MARK and FLAG support

Filter actions MARK and FLAG are supported on Medford2 by DPDK
firmware variant.

Submitted by:   Roman Zhukov <roman.zhukov at oktetlabs.ru>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18238

5 years agosfxge(4): add equal stride super-buffer prefix layout
Andrew Rybchenko [Thu, 29 Nov 2018 06:45:50 +0000 (06:45 +0000)]
sfxge(4): add equal stride super-buffer prefix layout

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18237

5 years agosfxge(4): support equal stride super-buffer Rx mode
Andrew Rybchenko [Thu, 29 Nov 2018 06:45:38 +0000 (06:45 +0000)]
sfxge(4): support equal stride super-buffer Rx mode

Equal stride super-buffer Rx mode is supported by DPDK firmware
variant. One Rx descriptor provides many Rx buffers to firmware.
Rx buffers follow each other with specified stride.
Also it supports head of line blocking with timeout to address
drops when no Rx descriptors are available. So it gives extra time
to the driver to provide Rx descriptors before drop.

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18236

5 years agosfxge(4): detect equal stride super-buffer support
Andrew Rybchenko [Thu, 29 Nov 2018 06:45:26 +0000 (06:45 +0000)]
sfxge(4): detect equal stride super-buffer support

Equal stride super-buffer Rx mode is supported on Medford2 by
DPDK firmware variant.

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18235

5 years agosfxge(4): make RxQ type data an union
Andrew Rybchenko [Thu, 29 Nov 2018 06:45:15 +0000 (06:45 +0000)]
sfxge(4): make RxQ type data an union

The type is an internal interface. Single integer is insufficient
to carry RxQ type-specific information in the case of equal stride
super-buffer Rx mode (packet buffers per bucket, maximum DMA length,
packet stride, head of line block timeout).

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18234

5 years agosfxge(4): update autogenerated MCDI and TLV headers
Andrew Rybchenko [Thu, 29 Nov 2018 06:45:04 +0000 (06:45 +0000)]
sfxge(4): update autogenerated MCDI and TLV headers

Equal stride super-buffer is a new name instead of deprecated equal
stride packed stream to avoid confusion with previous packed stream.

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18233

5 years agosfxge(4): honour packed stream RSS restriction
Andrew Rybchenko [Thu, 29 Nov 2018 06:44:51 +0000 (06:44 +0000)]
sfxge(4): honour packed stream RSS restriction

Packed stream firmware variant on EF10 adapters has a
number of properties which must be taken into account:

 - Only one exclusive RSS context is available per port.
 - Only IP addresses can contribute to the hash value.

Huntington and Medford have one more limitation which
is important for the drivers capable of packed stream:

 - Hash algorithm is non-standard (i.e. non-Toeplitz).
   This implies XORing together source + destination
   IP addresses (or last four bytes in the case of IPv6)
   and using the result as the input to a Toeplitz hash.

This patch provides a number of improvements in order
to treat the mentioned limitations in the common code.

If the firmware variant is packed stream, the list of
supported hash tuples will include less variants, and
the maximum number of RSS contexts will be set to one.

Submitted by:   Ivan Malov <ivan.malov at oktetlabs.ru>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18232

5 years agosfxge(4): support more RSS hash configurations
Andrew Rybchenko [Thu, 29 Nov 2018 06:44:39 +0000 (06:44 +0000)]
sfxge(4): support more RSS hash configurations

Modern firmwares on EF10 adapters have support for
more traffic classes eligible for hash computation.
Also, it has become possible to adjust hashing per
individual class and select distinct packet fields
which will be able to contribute to the hash value.

This patch adds support for the mentioned features.

Submitted by:   Ivan Malov <ivan.malov at oktetlabs.ru>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18231

5 years agosfxge(4): add a new means to control RSS hash
Andrew Rybchenko [Thu, 29 Nov 2018 06:44:28 +0000 (06:44 +0000)]
sfxge(4): add a new means to control RSS hash

Currently, libefx has no support for additional RSS modes
available with later controllers. In order to support this,
libefx should be able to list available hash configurations.

This patch provides basic infrastructure for the new interface.
The client drivers will be able to query the list of supported
hash configurations for a particular hash algorithm. Also, it
will be possible to configure hashing by means of new definitions.

Submitted by:   Ivan Malov <ivan.malov at oktetlabs.ru>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18230

5 years agosfxge(4): cope with clang warning on negative shift
Andrew Rybchenko [Thu, 29 Nov 2018 06:44:20 +0000 (06:44 +0000)]
sfxge(4): cope with clang warning on negative shift

clang 4.0.1-6 on Ubuntu generates false positive warning that shift
is negative.  It is done regardless of the fact that the branch is
not taken because of previous check.

The warning is generate in EFX_INSERT_NATIVE32 used by
EFX_INSERT_FIELD_NATIVE32. All similar cases are fixed as well.

It is undesirable to suppress the warning completely.

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18229

5 years agosfxge(4): fix comparison always true warning
Andrew Rybchenko [Thu, 29 Nov 2018 06:44:09 +0000 (06:44 +0000)]
sfxge(4): fix comparison always true warning

Loopback type used as bit index has efx_loopback_type_t type
which is enum. clang complains that it is always true when it
is compared with qword (64 bit) bits number boundary.

Sponsored by:   Solarflare Communications, Inc.
MFC after:      1 week
Differential Revision:  https://reviews.freebsd.org/D18228

5 years agosfxge(4): support FW subvariant choice
Andrew Rybchenko [Thu, 29 Nov 2018 06:43:57 +0000 (06:43 +0000)]
sfxge(4): support FW subvariant choice

If DPDK application or OS does not need checksumming on transmit,
it may be disabled in firmware to achieve higher packet rates.
Choice must be done before VIS allocation and is allowed if
no other non-preboot and firmware subvariant-unaware drivers are
attached.

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18227

5 years agosfxge(4): report no Tx checksum FW subvariant support
Andrew Rybchenko [Thu, 29 Nov 2018 06:43:46 +0000 (06:43 +0000)]
sfxge(4): report no Tx checksum FW subvariant support

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18226

5 years agosfxge(4): add firmware subvariant aware driver option
Andrew Rybchenko [Thu, 29 Nov 2018 06:43:34 +0000 (06:43 +0000)]
sfxge(4): add firmware subvariant aware driver option

FW subvariants allow to tweak NIC global features. For example,
if no drivers require checksumming on transmit, it may be disabled
in FW to increase packet rate.

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18225

5 years agosfxge(4): update MCDI headers
Andrew Rybchenko [Thu, 29 Nov 2018 06:43:23 +0000 (06:43 +0000)]
sfxge(4): update MCDI headers

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18224

5 years agosfxge(4): add values for RxDPCPU firmware id recognition
Andrew Rybchenko [Thu, 29 Nov 2018 06:43:12 +0000 (06:43 +0000)]
sfxge(4): add values for RxDPCPU firmware id recognition

Submitted by:   Roman Zhukov <roman.zhukov at oktetlabs.ru>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18223

5 years agosfxge(4): support choosing firmware variant
Andrew Rybchenko [Thu, 29 Nov 2018 06:43:00 +0000 (06:43 +0000)]
sfxge(4): support choosing firmware variant

Submitted by:   Gautam Dawar <gdawar at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18222

5 years agosfxge(4): support drop filters on EF10 family NICs
Andrew Rybchenko [Thu, 29 Nov 2018 06:42:49 +0000 (06:42 +0000)]
sfxge(4): support drop filters on EF10 family NICs

Add support for filters which drop packets when forming MCDI request
for a filter.

Submitted by:   Roman Zhukov <roman.zhukov at oktetlabs.ru>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18221

5 years agosfxge(4): distinguish filters for encapsulated packets
Andrew Rybchenko [Thu, 29 Nov 2018 06:42:38 +0000 (06:42 +0000)]
sfxge(4): distinguish filters for encapsulated packets

Add filter match flag to distinguish filters applied only to
encapsulated packets.

Match flags set should allow to determine whether a filter
is supported or not. The problem is that if specification
has supported set outer match flags and specified
encapsulation without any inner flags, check says that it
is supported, and filter insertion is performed. However,
there is no filtering of the encapsulated traffic. A new
flag is added to solve this problem and separate the
filters for the encapsulated packets.

Submitted by:   Roman Zhukov <roman.zhukov at oktetlabs.ru>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18220

5 years agosfxge(4): support VXLAN filter creation
Andrew Rybchenko [Thu, 29 Nov 2018 06:42:26 +0000 (06:42 +0000)]
sfxge(4): support VXLAN filter creation

Submitted by:   Vijay Srivastava <vijays at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18219

5 years agosfxge(4): support VNI/VSID and inner frame local MAC
Andrew Rybchenko [Thu, 29 Nov 2018 06:42:15 +0000 (06:42 +0000)]
sfxge(4): support VNI/VSID and inner frame local MAC

This supports VNI/VSID and inner frame local MAC fields to
match in VXLAN, GENEVE, or NVGRE packets.

Submitted by:   Roman Zhukov <roman.zhukov at oktetlabs.ru>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18218

5 years agosfxge(4): support filters for encapsulated packets
Andrew Rybchenko [Thu, 29 Nov 2018 06:42:04 +0000 (06:42 +0000)]
sfxge(4): support filters for encapsulated packets

This adds filters for encapsulated packets to the list
returned by ef10_filter_supported_filters().

Submitted by:   Roman Zhukov <roman.zhukov at oktetlabs.ru>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18217

5 years agosfxge(4): sync MCDI headers and TLV layout
Andrew Rybchenko [Thu, 29 Nov 2018 06:41:53 +0000 (06:41 +0000)]
sfxge(4): sync MCDI headers and TLV layout

Regenerate MCDI and TLV layout headers from firmwaresrc to
pick up DPDK firmware variant and related Rx queue and filtering
extensions.

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18216

5 years agosfxge(4): add signed image layout support
Andrew Rybchenko [Thu, 29 Nov 2018 06:41:41 +0000 (06:41 +0000)]
sfxge(4): add signed image layout support

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18215

5 years agosfxge(4): add firmware image layout option
Andrew Rybchenko [Thu, 29 Nov 2018 06:29:24 +0000 (06:29 +0000)]
sfxge(4): add firmware image layout option

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18214

5 years agoConvert racct_enable to bool and annotate as __read_frequently
Mateusz Guzik [Thu, 29 Nov 2018 05:17:16 +0000 (05:17 +0000)]
Convert racct_enable to bool and annotate as __read_frequently

Sponsored by: The FreeBSD Foundation

5 years agoDeinline racct throttling out of syscall exit path.
Mateusz Guzik [Thu, 29 Nov 2018 05:08:46 +0000 (05:08 +0000)]
Deinline racct throttling out of syscall exit path.

racct is not enabled by default and even when it is enabled processes are
typically not throttled. The order of checks is left unchanged since
racct_enable will be annotated as __read_frequently, while checking for the
flag in the processes would probably require an extra fetch.

Sponsored by: The FreeBSD Foundation

5 years agoAnnotate td_cowgen check as unlikely.
Mateusz Guzik [Thu, 29 Nov 2018 04:48:22 +0000 (04:48 +0000)]
Annotate td_cowgen check as unlikely.

Sponsored by: The FreeBSD Foundation

5 years agoAllow truss to operate on ELFv2 processes.
Justin Hibbits [Thu, 29 Nov 2018 03:57:06 +0000 (03:57 +0000)]
Allow truss to operate on ELFv2 processes.

Summary:
Currently, truss doesn't work on ELFv2 processes due to not recognizing the ABI.

Since there's no special handling needed here, just adding a PROCABI struct for
it is sufficient to get it working.

Submitted by: git_bdragon.rtk0.net
Differential Revision: https://reviews.freebsd.org/D18352

5 years agoTidy up hardclock.
Mateusz Guzik [Thu, 29 Nov 2018 03:44:02 +0000 (03:44 +0000)]
Tidy up hardclock.

- use fcmpset for updating ticks
- move (rarely used) itimer handling to a dedicated function

Sponsored by: The FreeBSD Foundation

5 years agoFix thread creation in PowerPC64 ELFv2 processes.
Justin Hibbits [Thu, 29 Nov 2018 03:39:11 +0000 (03:39 +0000)]
Fix thread creation in PowerPC64 ELFv2 processes.

Summary:
Currently, the upcall used to create threads assumes ELFv1.

Instead, we should check which sysentvec is in use on the process and act
accordingly.

This makes ELFv2 threaded processes work.

Submitted by: git_bdragon.rtk0.net
Differential Revision: https://reviews.freebsd.org/D18330

5 years agoproc: create a dedicated lock for zombproc to ligthen the load on allproc_lock
Mateusz Guzik [Thu, 29 Nov 2018 02:52:08 +0000 (02:52 +0000)]
proc: create a dedicated lock for zombproc to ligthen the load on allproc_lock

waitpid always takes proctree to evaluate the list, but only takes allproc
if it can reap. With this patch allproc is no longer taken, which helps during
poudriere -j 128.

Discussed with: kib
Sponsored by: The FreeBSD Foundation

5 years agoMove CLIP table handling out of TOM and into the base driver.
John Baldwin [Thu, 29 Nov 2018 01:15:53 +0000 (01:15 +0000)]
Move CLIP table handling out of TOM and into the base driver.

- Store the clip table in 'struct adapter' instead of in the TOM softc.
- Init the clip table during attach and teardown during detach.
- While here, add a dev.<nexus>.<unit>.misc.clip sysctl to dump the
  CLIP table.

This does mean that we update the clip table even if TOE is not enabled,
but non-TOE things need the CLIP table anyway.

Reviewed by: np, Krishnamraju Eraparaju @ Chelsio
Sponsored by: Chelsio Communications
Differential Revision: https://reviews.freebsd.org/D18010

5 years agosh: Fix ${param?} default error message
Jilles Tjoelker [Wed, 28 Nov 2018 20:03:53 +0000 (20:03 +0000)]
sh: Fix ${param?} default error message

If word in ${param?word} is missing, the shell shall write a default error
message. So expanding ${param?} when param is not set should write an error
message like

sh: param: parameter not set

This was broken by r316417.

PR: 233585

5 years agoFix yet another edge case in uma_startup_count(). If zone size fits into
Gleb Smirnoff [Wed, 28 Nov 2018 19:54:02 +0000 (19:54 +0000)]
Fix yet another edge case in uma_startup_count(). If zone size fits into
several pages, but leaves no space for struct uma_slab at the end we
miscalculate number of pages by one. Totally mimic keg_large_init() math
here to cover that problem.

Reported by: gallatin

5 years agoFor not offpage zones the slab is placed at the end of page. Keg's uk_pgoff
Gleb Smirnoff [Wed, 28 Nov 2018 19:17:27 +0000 (19:17 +0000)]
For not offpage zones the slab is placed at the end of page. Keg's uk_pgoff
is calculated to guarantee that struct uma_slab is placed at pointer size
alignment. Calculation of real struct uma_slab size is done in keg_ctor()
and yet again in keg_large_init(), to check if we need an extra page. This
calculation can actually be performed at compile time.

- Add SIZEOF_UMA_SLAB macro to calculate size of struct uma_slab placed at
  an end of a page with alignment requirement.
- Use SIZEOF_UMA_SLAB in keg_ctor() and in keg_large_init(). This is a not
  a functional change.
- Use SIZEOF_UMA_SLAB in UMA_SLAB_SPACE definition and in keg_small_init().
  This is a potential bugfix, but in reality I don't think there are any
  systems affected, since compiler aligns struct uma_slab anyway.

5 years agoAdd comments describing the bootargs handoff between loader(8) and gptboot
Ian Lepore [Wed, 28 Nov 2018 18:09:42 +0000 (18:09 +0000)]
Add comments describing the bootargs handoff between loader(8) and gptboot
or zfsboot, when loader(8) is the BTX loader.  No functional changes.

5 years agoFix first-packet completion
Stephen Hurd [Wed, 28 Nov 2018 17:37:17 +0000 (17:37 +0000)]
Fix first-packet completion

The first packet after the ring is initialized was never
completed as isc_txd_credits_update() would not include it in the
count of completed packets. This caused netmap to never complete
a batch. See PR 233022 for more details.

This is the same fix as the r340310 for e1000

PR: 233607
Reported by: lev
Reviewed by: lev
MFC after: 3 days
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D18368

5 years agopowerpc: Don't include KERNBASE in genassym, it's unnecessary
Justin Hibbits [Wed, 28 Nov 2018 16:00:52 +0000 (16:00 +0000)]
powerpc: Don't include KERNBASE in genassym, it's unnecessary

A related future change, which changes KERNBASE for Book-E for some reason
causes a "KERNBASE redefined" error with assym.inc, even though it only changed
the value of KERNBASE and nothing else.  Since machine/vmparam.h is already
included in booke/locore.S, and the requisite guards are already in place for
properly handling KERNBASE in vmparam.h, just remove it from genassym, and
include vmparam.h in the AIM locore files.

5 years agocxgbe: revert r309725
Vincenzo Maffione [Wed, 28 Nov 2018 15:29:58 +0000 (15:29 +0000)]
cxgbe: revert r309725

After the fix contained in r341144, cxgbe does not need anymore
to set the IFCAP_NETMAP flag manually.

Reviewed by: np
Approved by: gnn (mentor)
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D17987

5 years agonetmap: set IFCAP_NETMAP in if_capabilities
Vincenzo Maffione [Wed, 28 Nov 2018 14:07:34 +0000 (14:07 +0000)]
netmap: set IFCAP_NETMAP in if_capabilities

Revision r307394 removed (by mistake) the code that sets IFCAP_NETMAP
in if_capabilities on netmap_attach. This patch reverts this change.

Reviewed by: np
Approved by: gnn (mentor)
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D17987

5 years agoarm64: Add rk_i2c and rk805 to GENERIC
Emmanuel Vadot [Wed, 28 Nov 2018 13:54:39 +0000 (13:54 +0000)]
arm64: Add rk_i2c and rk805 to GENERIC

MFC after: 1 month

5 years agoAdd RK805 PMIC Support
Emmanuel Vadot [Wed, 28 Nov 2018 13:53:43 +0000 (13:53 +0000)]
Add RK805 PMIC Support

RK805 is the companion PMIC for RK3328 SoC.
Add a driver for it with most of it's regulators supported.

MFC after: 1 month

5 years agork_clk_composite: Fix set_mux
Emmanuel Vadot [Wed, 28 Nov 2018 13:25:22 +0000 (13:25 +0000)]
rk_clk_composite: Fix set_mux

Pointy Hat:   manu

5 years agoAdd CAPABILITIES to the ERL kernel config so that tools that have been
Sean Bruno [Wed, 28 Nov 2018 13:25:10 +0000 (13:25 +0000)]
Add CAPABILITIES to the ERL kernel config so that tools that have been
modified with Capsicum work on this target platform.

This came up after the conversion of wc(8).

5 years agosfxge(4): add outer IP ID parameter to TSOv2 descriptor
Andrew Rybchenko [Wed, 28 Nov 2018 09:25:43 +0000 (09:25 +0000)]
sfxge(4): add outer IP ID parameter to TSOv2 descriptor

Set outer_ip_id in the TX option descriptor for encapsulated packets.

Submitted by:   Vijay Srivastava <vijays at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18213

5 years agosfxge(4): add encapsulated TSOv2 capability
Andrew Rybchenko [Wed, 28 Nov 2018 09:25:31 +0000 (09:25 +0000)]
sfxge(4): add encapsulated TSOv2 capability

Submitted by:   Vijay Srivastava <vijays at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18212

5 years agosfxge(4): support CTPIO stats
Andrew Rybchenko [Wed, 28 Nov 2018 09:25:21 +0000 (09:25 +0000)]
sfxge(4): support CTPIO stats

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18211

5 years agosfxge(4): regenerate headers to pick up CTPIO stats
Andrew Rybchenko [Wed, 28 Nov 2018 09:25:09 +0000 (09:25 +0000)]
sfxge(4): regenerate headers to pick up CTPIO stats

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18210

5 years agosfxge(4): add bit to indicate CTPIO availability
Andrew Rybchenko [Wed, 28 Nov 2018 09:24:58 +0000 (09:24 +0000)]
sfxge(4): add bit to indicate CTPIO availability

Submitted by:   Guido Barzini <gbarzini at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18209

5 years agosfxge(4): add CTPIO statistics
Andrew Rybchenko [Wed, 28 Nov 2018 09:24:47 +0000 (09:24 +0000)]
sfxge(4): add CTPIO statistics

Submitted by:   Guido Barzini <gbarzini at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18208

5 years agosfxge(4): run genfwdef to update headers
Andrew Rybchenko [Wed, 28 Nov 2018 09:24:36 +0000 (09:24 +0000)]
sfxge(4): run genfwdef to update headers

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18207

5 years agosfxge(4): document the event type for CTPIO sends
Andrew Rybchenko [Wed, 28 Nov 2018 09:24:25 +0000 (09:24 +0000)]
sfxge(4): document the event type for CTPIO sends

Document the TX_EV_TYPE used for TX completion events corresponding
to CTPIO sends.

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18206

5 years agosfxge(4): add 1.3V voltage and current sensors
Andrew Rybchenko [Wed, 28 Nov 2018 09:24:14 +0000 (09:24 +0000)]
sfxge(4): add 1.3V voltage and current sensors

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18205

5 years agosfxge(4): provide a flag for controlling CTPIO mode
Andrew Rybchenko [Wed, 28 Nov 2018 09:24:03 +0000 (09:24 +0000)]
sfxge(4): provide a flag for controlling CTPIO mode

Either cut-through or store-and-forward mode.

Submitted by:   Guido Barzini <gbarzini at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18204

5 years agosfxge(4): add Medford2 support for tunnel encapsulations
Andrew Rybchenko [Wed, 28 Nov 2018 09:23:52 +0000 (09:23 +0000)]
sfxge(4): add Medford2 support for tunnel encapsulations

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18203

5 years agosfxge(4): add Medford2 support for licensing
Andrew Rybchenko [Wed, 28 Nov 2018 09:23:41 +0000 (09:23 +0000)]
sfxge(4): add Medford2 support for licensing

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18202

5 years agosfxge(4): add Medford2 support for external port numbers
Andrew Rybchenko [Wed, 28 Nov 2018 09:23:30 +0000 (09:23 +0000)]
sfxge(4): add Medford2 support for external port numbers

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18201

5 years agosfxge(4): group Medford external port mapping entries
Andrew Rybchenko [Wed, 28 Nov 2018 09:23:19 +0000 (09:23 +0000)]
sfxge(4): group Medford external port mapping entries

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18200

5 years agosfxge(4): clarify port mode names and masks
Andrew Rybchenko [Wed, 28 Nov 2018 09:23:05 +0000 (09:23 +0000)]
sfxge(4): clarify port mode names and masks

New port mode names are defined for Medford2 and later, and
the existing names are aliased to them. Add comments with the
numeric port mode to clarify the external port modes table.

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18199

5 years agosfxge(4): support Medford2 event timer semantics
Andrew Rybchenko [Wed, 28 Nov 2018 09:22:53 +0000 (09:22 +0000)]
sfxge(4): support Medford2 event timer semantics

The event timer interface has changed for Medford2 - for
details see bug66418 comment 9. Update the common code to
use the new timer semantics for Medford2.

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18198

5 years agosfxge(4): support FEC mode settings
Andrew Rybchenko [Wed, 28 Nov 2018 09:22:42 +0000 (09:22 +0000)]
sfxge(4): support FEC mode settings

Medford2 controllers support control and reporting of
FEC modes for 25G and higher links. See SF-109306-TC
for suggested usage in client code.

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18197

5 years agosfxge(4): remove obsolete comments
Andrew Rybchenko [Wed, 28 Nov 2018 06:56:34 +0000 (06:56 +0000)]
sfxge(4): remove obsolete comments

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18196

5 years agosfxge(4): move VI window size config to ef10 NIC board
Andrew Rybchenko [Wed, 28 Nov 2018 06:56:22 +0000 (06:56 +0000)]
sfxge(4): move VI window size config to ef10 NIC board

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18195

5 years agosfxge(4): move privilege config to ef10 NIC board config
Andrew Rybchenko [Wed, 28 Nov 2018 06:56:11 +0000 (06:56 +0000)]
sfxge(4): move privilege config to ef10 NIC board config

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18194

5 years agosfxge(4): move vector config to ef10 NIC board config
Andrew Rybchenko [Wed, 28 Nov 2018 06:55:59 +0000 (06:55 +0000)]
sfxge(4): move vector config to ef10 NIC board config

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18193

5 years agosfxge(4): move limits config to ef10 NIC board config
Andrew Rybchenko [Wed, 28 Nov 2018 06:55:47 +0000 (06:55 +0000)]
sfxge(4): move limits config to ef10 NIC board config

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18192

5 years agosfxge(4): move Tx config to ef10 NIC board config
Andrew Rybchenko [Wed, 28 Nov 2018 06:55:36 +0000 (06:55 +0000)]
sfxge(4): move Tx config to ef10 NIC board config

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18191

5 years agosfxge(4): move Rx config to ef10 NIC board config
Andrew Rybchenko [Wed, 28 Nov 2018 06:55:24 +0000 (06:55 +0000)]
sfxge(4): move Rx config to ef10 NIC board config

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18190

5 years agosfxge(4): move datapath config to ef10 NIC board cfg
Andrew Rybchenko [Wed, 28 Nov 2018 06:55:12 +0000 (06:55 +0000)]
sfxge(4): move datapath config to ef10 NIC board cfg

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18189

5 years agosfxge(4): move PHY/link config to ef10 NIC board cfg
Andrew Rybchenko [Wed, 28 Nov 2018 06:55:00 +0000 (06:55 +0000)]
sfxge(4): move PHY/link config to ef10 NIC board cfg

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18188

5 years agosfxge(4): move legacy board config to ef10 NIC board cfg
Andrew Rybchenko [Wed, 28 Nov 2018 06:54:49 +0000 (06:54 +0000)]
sfxge(4): move legacy board config to ef10 NIC board cfg

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18187

5 years agosfxge(4): move MAC address config to ef10 NIC board cfg
Andrew Rybchenko [Wed, 28 Nov 2018 06:54:37 +0000 (06:54 +0000)]
sfxge(4): move MAC address config to ef10 NIC board cfg

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18186

5 years agosfxge(4): move PF/VF config to ef10 NIC board config
Andrew Rybchenko [Wed, 28 Nov 2018 06:54:25 +0000 (06:54 +0000)]
sfxge(4): move PF/VF config to ef10 NIC board config

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18185

5 years agosfxge(4): move port config to ef10 NIC board config
Andrew Rybchenko [Wed, 28 Nov 2018 06:54:13 +0000 (06:54 +0000)]
sfxge(4): move port config to ef10 NIC board config

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18184

5 years agosfxge(4): pacify false positive warning
Andrew Rybchenko [Wed, 28 Nov 2018 06:54:02 +0000 (06:54 +0000)]
sfxge(4): pacify false positive warning

The warning says that it may be usage of possibly uninitialized value.

Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18183

5 years agosfxge(4): add ef10 NIC board config method
Andrew Rybchenko [Wed, 28 Nov 2018 06:53:51 +0000 (06:53 +0000)]
sfxge(4): add ef10 NIC board config method

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18182

5 years agosfxge(4): use correct name for frame truncation event
Andrew Rybchenko [Wed, 28 Nov 2018 06:53:40 +0000 (06:53 +0000)]
sfxge(4): use correct name for frame truncation event

The RX_ECC_ERR flag in RX events was misnamed, as it
reported frame truncation. Use the new RX_TRUNC_ERR
name for this flag.

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18181

5 years agopowerpc/booke: Fix debug printfs in pmap
Justin Hibbits [Wed, 28 Nov 2018 04:02:26 +0000 (04:02 +0000)]
powerpc/booke: Fix debug printfs in pmap

Add missing '%'s so printf formats are actually handled.

5 years agopowerpc: Fix the powerpc64 build post-r341102
Justin Hibbits [Wed, 28 Nov 2018 02:48:43 +0000 (02:48 +0000)]
powerpc: Fix the powerpc64 build post-r341102

VM_MIN_KERNEL_ADDRESS is now used in locore.S, but the UL suffix isn't
permitted in .S files.

5 years agopowerpc: Prepare Book-E kernels for KERNBASE != run base
Justin Hibbits [Wed, 28 Nov 2018 02:00:27 +0000 (02:00 +0000)]
powerpc: Prepare Book-E kernels for KERNBASE != run base

Book-E kernels really run at VM_MIN_KERNEL_ADDRESS, which currently happens to
be the same as KERNBASE.  KERNBASE is the linked address, which the loader also
takes to be the physical load address.  Treat KERNBASE as a physical address,
not a virtual, and change virtual address references for KERNBASE to use
something more appropriate.

5 years agopowerpcspe: Don't crash the loader on ubldr with SPE instructions.
Justin Hibbits [Wed, 28 Nov 2018 01:47:01 +0000 (01:47 +0000)]
powerpcspe: Don't crash the loader on ubldr with SPE instructions.

-msoft-float seems to be insufficient for disabling the SPE on powerpcspe.
Force it off with -mno-spe as well.  This prevents a crash in ubldr on
powerpcspe.

5 years agoAdd read-only sysctls for all tunables in the cxgbe(4) driver.
John Baldwin [Tue, 27 Nov 2018 22:02:54 +0000 (22:02 +0000)]
Add read-only sysctls for all tunables in the cxgbe(4) driver.

Reviewed by: np
MFC after: 1 month
Sponsored by: Chelsio Communications
Differential Revision: https://reviews.freebsd.org/D18360

5 years agosh: Use 126 and 127 exit status for failures opening a script
Jilles Tjoelker [Tue, 27 Nov 2018 21:49:59 +0000 (21:49 +0000)]
sh: Use 126 and 127 exit status for failures opening a script

This affects scripts named on the command line, named with a '.' special
builtin and found via the PATH %func autoloading mechanism.

PR: 231986

5 years agoFix assert condition in pmap_large_unmap().
Konstantin Belousov [Tue, 27 Nov 2018 21:40:51 +0000 (21:40 +0000)]
Fix assert condition in pmap_large_unmap().

pmap_large_unmap() asserts that an unmapping request covers the
entirety of a 2M or 1G page.  The logic in the asserts was out of date
with the loop logic.  Correct the test to actually check that
destroying the current superpage mapping does not unmap addresses
beyond those requested by the caller.

Submitted by: D Scott Phillips <d.scott.phillips@intel.com>
Reviewed by: alc
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D18345

5 years agoUse busdma unconditionally in iflib
Andrew Gallatin [Tue, 27 Nov 2018 20:01:05 +0000 (20:01 +0000)]
Use busdma unconditionally in iflib

- Remove the complex mechanism to choose between using busdma
and raw pmap_kextract at runtime.   The reduced complexity makes
the code easier to read and maintain.

- Fix a bug in the small packet receive path where clusters were
repeatedly mapped but never unmapped. We now store the cluster's
bus address and avoid re-mapping the cluster each time a small
packet is received.

This patch fixes bugs I've seen where ixl(4) will not even
respond to ping without seeing DMAR faults.

I see a small improvement (14%) on packet forwarding tests using
a Haswell based Xeon E5-2697 v3.  Olivier sees a small
regression (-3% to -6%) with lower end hardware.

Reviewed by: mmacy
Not objected to by: sbruno
MFC after: 8 weeks
Sponsored by: Netflix, Inc
Differential Revision: https://reviews.freebsd.org/D17901

5 years agoImprove sigonstack().
Konstantin Belousov [Tue, 27 Nov 2018 19:50:58 +0000 (19:50 +0000)]
Improve sigonstack().

Avoid relying on unsigned overflow for the test.
Simplify expressions to avoid duplicate check for the range.
Style.
Add herald comment.

Reviewed by: jhb
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D18361

5 years agoIn hardened systems, where the security.bsd.unprivileged_proc_debug sysctl
Jamie Gritton [Tue, 27 Nov 2018 17:51:50 +0000 (17:51 +0000)]
In hardened systems, where the security.bsd.unprivileged_proc_debug sysctl
node is set, allow setting security.bsd.unprivileged_proc_debug per-jail.
In part, this is needed to create jails in which the Address Sanitizer
(ASAN) fully works as ASAN utilizes libkvm to inspect the virtual address
space. Instead of having to allow unprivileged process debugging for the
entire system, allow setting it on a per-jail basis.

The sysctl node is still security.bsd.unprivileged_proc_debug and the
jail(8) param is allow.unprivileged_proc_debug. The sysctl code is now a
sysctl proc rather than a sysctl int. This allows us to determine setting
the flag for the corresponding jail (or prison0).

As part of the change, the dynamic allow.* API needed to be modified to
take into account pr_allow flags which may now be disabled in prison0.
This prevents conflicts with new pr_allow flags (like that of vmm(4)) that
are added (and removed) dynamically.

Also teach the jail creation KPI to allow differences for certain pr_allow
flags between the parent and child jail. This can happen when unprivileged
process debugging is disabled in the parent prison, but enabled in the
child.

Submitted by: Shawn Webb <lattera at gmail.com>
Obtained from: HardenedBSD (45b3625edba0f73b3e3890b1ec3d0d1e95fd47e1deba0b5078cef0faae43cbdafed3035b16587afcab21eeb3b4c72f2500987c96ff603ccf3b6e7de8)
Relnotes: yes
Sponsored by: HardenedBSD and G2, Inc
Differential Revision: https://reviews.freebsd.org/D18319

5 years agoDo not limit the mbuf queue length for keepalive packets.
Andrey V. Elsukov [Tue, 27 Nov 2018 16:51:01 +0000 (16:51 +0000)]
Do not limit the mbuf queue length for keepalive packets.

It was unlimited before overhaul, and one user reported that this limit
can be reached easily.

PR: 233562
MFC after: 1 week

5 years agoWhen deciding whether to send the complete URL or just the document part,
Dag-Erling Smørgrav [Tue, 27 Nov 2018 16:23:17 +0000 (16:23 +0000)]
When deciding whether to send the complete URL or just the document part,
we were looking at the original URL rather than the one we were currently
processing.  This meant that if we were trying to retrieve an HTTP URL but
were redirected to an HTTPS URL, and HTTPS proxying was enabled, we would
send an invalid request and most likely get garbage back.

MFC after: 3 days

5 years agoRestore the ability to override the disk unit/partition at the boot: prompt
Ian Lepore [Tue, 27 Nov 2018 16:16:38 +0000 (16:16 +0000)]
Restore the ability to override the disk unit/partition at the boot: prompt
in gptboot.

When arch-independent geli support was added, a new static 'gdsk' struct
was added, but there was still a static 'dsk' struct, and when you typed
in an alternate disk/partition, the string was parsed into that struct,
which was then never used for anything.  Now the string gets parsed into
gdsk.dsk, the struct that's actually used.

X-MFC after: 3 days

5 years agoMake whitespace more consistent in libkqueue tests.
David Bright [Tue, 27 Nov 2018 15:12:34 +0000 (15:12 +0000)]
Make whitespace more consistent in libkqueue tests.

After r337820, which "corrected" some spaces-instead-of-tab whitespace
issues in the libkqueue tests, jmg@ pointed out that these files were
originally space-based, not tab-spaced, and so the correction should
have been to get rid of the tabs that had been introduced in previous
changes, not the spaces. This change does that. This is a whitespace
only change; no functional change is intended.

Reported by: jmg@
MFC after: 3 days
Sponsored by: Dell EMC Isilon

5 years agoAdd a “skip_dsn” option to g_part's bootcode verb to prevent g_part_mbr
Dag-Erling Smørgrav [Tue, 27 Nov 2018 14:58:19 +0000 (14:58 +0000)]
Add a “skip_dsn” option to g_part's bootcode verb to prevent g_part_mbr
from setting the volume serial number.  This unbreaks older boot blocks
that don't support serial numbers, and allows boot0cfg to set the serial
number itself if requested by the user.

Submitted by: lev@, yuripv@
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D17386

5 years agosfxge(4): remove MAC stats size define
Andrew Rybchenko [Tue, 27 Nov 2018 14:16:14 +0000 (14:16 +0000)]
sfxge(4): remove MAC stats size define

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18180

5 years agosfxge(4): resolve code analysis warnings
Andrew Rybchenko [Tue, 27 Nov 2018 14:16:03 +0000 (14:16 +0000)]
sfxge(4): resolve code analysis warnings

Minimal changes adding buffer size checks and simplifying checksum
processing.

Submitted by:   Richard Houldsworth <rhouldsworth at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18179

5 years agosfxge(4): decode Medford2 FEC stats if available
Andrew Rybchenko [Tue, 27 Nov 2018 14:15:52 +0000 (14:15 +0000)]
sfxge(4): decode Medford2 FEC stats if available

Decode Medford2 FEC stats if available in MAC stats DMA buffer.

Submitted by:   Andy Moreton <amoreton at solarflare.com>
Sponsored by:   Solarflare Communications, Inc.
Differential Revision:  https://reviews.freebsd.org/D18178