Add missing `;`

Approved by: kevlo

Support attaching tx queues to cpus

This will attempt to use a different thread/core on the same L2
cache when possible, or use the same cpu as the rx thread when not.
If SMP isn't enabled, don't go

Support attaching tx queues to cpus

This will attempt to use a different thread/core on the same L2
cache when possible, or use the same cpu as the rx thread when not.
If SMP isn't enabled, don't go looking for cores to use. This is mostly
useful when using shared TX/RX queues.

Reviewed by: sbruno
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D12446

show more ...

Update Matthew Macy contact info

Email address has changed, uses consistent name (Matthew, not Matt)

Reported by: Matthew Macy <mmacy@mattmacy.io>
Differential Revision: https://reviews.freebsd.org

Update Matthew Macy contact info

Email address has changed, uses consistent name (Matthew, not Matt)

Reported by: Matthew Macy <mmacy@mattmacy.io>
Differential Revision: https://reviews.freebsd.org/D13537

show more ...

Increment encap_pad_mbuf_fail when m_dup() fails in padding

Previously, the counter was only incremented when m_append() failed. Since
the function can also fail on m_dup() now, increment the count

Increment encap_pad_mbuf_fail when m_dup() fails in padding

Previously, the counter was only incremented when m_append() failed. Since
the function can also fail on m_dup() now, increment the counter there as
well.

Sponsored by: Limelight Networks

show more ...

Free mbuf chain when m_dup fails

Fix memory leak where mbuf chain wasn't free()d if iflib_ether_pad()
has a failure in m_dup().

Reported by: "Ryan Stone" <rysto32@gmail.com>
Sponsored by: Limelight

Free mbuf chain when m_dup fails

Fix memory leak where mbuf chain wasn't free()d if iflib_ether_pad()
has a failure in m_dup().

Reported by: "Ryan Stone" <rysto32@gmail.com>
Sponsored by: Limelight Networks

show more ...

Handle read-only mbufs in iflib ether pad function

If ethernet padding is enabled, and a read-only mbuf is passed,
it would modify the mbuf using m_append(). Instead, call m_dup() and
append to the

Handle read-only mbufs in iflib ether pad function

If ethernet padding is enabled, and a read-only mbuf is passed,
it would modify the mbuf using m_append(). Instead, call m_dup() and
append to the new packet.

Reported by: Pyun YongHyeon
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D13414

show more ...

iflib: Support to padding Ethernet frames to a min size

Some bnxt devices do not correctly send frames smaller than
52 bytes (without CRC), so add a quirk that will pad frames to an
arbitrary size b

iflib: Support to padding Ethernet frames to a min size

Some bnxt devices do not correctly send frames smaller than
52 bytes (without CRC), so add a quirk that will pad frames to an
arbitrary size before passing off to the encap routine.

Reported by: Bhargava Chenna Marreddy <bhargava.marreddy@broadcom.com>
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D13269

show more ...

Avoid calling CURVNET_[SET|RESTORE] for each packet

The LRO possible test was calling CURVNET_SET once for IPv4 or IPv6 for
each packet in a chain. Only call it once per chain instead.

Submitted by

Avoid calling CURVNET_[SET|RESTORE] for each packet

The LRO possible test was calling CURVNET_SET once for IPv4 or IPv6 for
each packet in a chain. Only call it once per chain instead.

Submitted by: Matthew Macy <mmacy@mattmacy.io>
Reviewed by: cem, ae
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D13368

show more ...

Add support for SIOCGIFXMEDIA to iflib

SIOCGIFXMEDIA is required for extended ethernet media types,
but iflib did not support it.

Reported by: Bhargava Chenna Marreddy <bhargava.marreddy@broadcom.c

Add support for SIOCGIFXMEDIA to iflib

SIOCGIFXMEDIA is required for extended ethernet media types,
but iflib did not support it.

Reported by: Bhargava Chenna Marreddy <bhargava.marreddy@broadcom.com>
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D13312

show more ...

Fix comment introduced in r326369

The code uses the set of all CPUs, it doesn't zero out the set.

Sponsored by: Limelight Networks

Ensure that ctx->ifc_cpus is always initialized

If a device didn't support MSI-X, ctx->ifc_cpus would not be initialized,
but the IRQ allocation routines still uses the value. Move the
initializati

Ensure that ctx->ifc_cpus is always initialized

If a device didn't support MSI-X, ctx->ifc_cpus would not be initialized,
but the IRQ allocation routines still uses the value. Move the
initialization to common code.

Sponsored by: Limelight Networks

show more ...

Merge ^/head r325999 through r326131.

Fix off-by-one error in bit_nclear() usage

bit_nclear() takes the bit numbers for the start and end bits, not the start
and a count. This was resulting in memory corruption past the end of the
bits

Fix off-by-one error in bit_nclear() usage

bit_nclear() takes the bit numbers for the start and end bits, not the start
and a count. This was resulting in memory corruption past the end of the
bitstr_t.

Sponsored by: Limelight Networks

show more ...

Merge ^/head r325842 through r325998.

Fix default numbers of iflib queue sets

The intent appears to be having one RX/TX queue set per core,
but since scctx->isc_n[tr]xqsets is set to max before calling
iflib_msix_init(), both end up bei

Fix default numbers of iflib queue sets

The intent appears to be having one RX/TX queue set per core,
but since scctx->isc_n[tr]xqsets is set to max before calling
iflib_msix_init(), both end up being set to total number of cores.

Use ctx->ifc_sysctl_n[rt]xqs as the selected value and
scctx->isc_n[rt]xqsets as the max. This should result in what appears
to be the intended behaviour

Reviewed by: sbruno
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D13096

show more ...

Merge ^/head r323559 through r325504.

Fix NOINET/NOINET6 build during compilation of iflib.

Reported by: kib

Only chain non-LRO mbufs when LRO is not possible

Preserve packet order between tcp_lro_rx() and if_input() to avoid
creating extra corner cases. If no packets can be LROed, combine them
into one ch

Only chain non-LRO mbufs when LRO is not possible

Preserve packet order between tcp_lro_rx() and if_input() to avoid
creating extra corner cases. If no packets can be LROed, combine them
into one chain for submission via if_input(). If any packet can
potentially be LROed however, retain old behaviour and call if_input()
for each packet.

This should keep the 12% improvement for small packet forwarding intact,
but mostly avoids impacting the LRO case.

Reviewed by: cem, sbruno
Approved by: sbruno (mentor)
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D12876

show more ...

MFhead@r325383

Preserve TSO checksum flags

r323941 incorrectly disabled TSO flags based on MTU.

Reported by: Yuri Pankov <yuripv@gmx.com>
Reviewed by: sbruno
Approved by: sbruno (mentor)
Sponsored by: Limelight N

Preserve TSO checksum flags

r323941 incorrectly disabled TSO flags based on MTU.

Reported by: Yuri Pankov <yuripv@gmx.com>
Reviewed by: sbruno
Approved by: sbruno (mentor)
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D12880

show more ...

Fix PR221990 - Assertion at iflib.c:1947

ifl_pidx and ifl_credits are going out of sync in _iflib_fl_refill() as they
use different update log. Use the same update logic for both, and add a
final c

Fix PR221990 - Assertion at iflib.c:1947

ifl_pidx and ifl_credits are going out of sync in _iflib_fl_refill() as they
use different update log. Use the same update logic for both, and add a
final call to isc_rxd_refill() to handle early exits from the loop.

PR: 221990
Reported by: pho
Reviewed by: sbruno
Approved by: sbruno (mentor)
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D12798

show more ...

MFhead@r325209

Fix build with nodevice netmap

iru_init() was declared and used outside the DEV_NETMAP
conditional blocks, but was implemented inside one. Move the
implementation out of the DEV_NETMAP block to allo

Fix build with nodevice netmap

iru_init() was declared and used outside the DEV_NETMAP
conditional blocks, but was implemented inside one. Move the
implementation out of the DEV_NETMAP block to allow building with
netmap disabled.

Reported by: Andrew Turner <andrew@fubar.geek.nz>
Reviewed by: sbruno
Approved by: sbruno (mentor)
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D12842

show more ...

MFhead@r325199

bnxt: HW_LRO Rx Pkt with > 32 fragments caused Crash (iflib)

Broadcom NIC with HW_LRO setting max_agg_segs >= 6 can generate Rx pkt with
64 (2^6) fragments, modify IFLIB_MAX_RX_SEGS to 64 to avoid m

bnxt: HW_LRO Rx Pkt with > 32 fragments caused Crash (iflib)

Broadcom NIC with HW_LRO setting max_agg_segs >= 6 can generate Rx pkt with
64 (2^6) fragments, modify IFLIB_MAX_RX_SEGS to 64 to avoid memory
corruption / Crash.

Submitted by: Bhargava Chenna Marreddy <bhargava.marreddy@broadcom.com>
Reviewed by: shurd, sbruno
Approved by: sbruno (mentor)
Sponsored by: Broadcom Limited
Differential Revision: https://reviews.freebsd.org/D12774

show more ...