Doing a 'dd' over iscsi will reliably cause stalls. Tx
cleaning _should_ reliably happen as data is sent.
However, currently if the transmit queue fills it will
wait until the iflib timer (hz/2) runs.
This change causes the the tx taskq thread to be run
if there are completed descriptors.
While here:
- make timer interrupt delay a sysctl
- simplify txd_db_check handling
- comment on INTR types
Background on the change:
Initially doorbell updates were minimized by only writing to the register
on every fourth packet. If txq_drain would return without writing to the
doorbell it scheduled a callout on the next tick to do the doorbell write
to ensure that the write otherwise happened "soon". At that time a sysctl
was added for users to avoid the potential added latency by simply writing
to the doorbell register on every packet. This worked perfectly well for
e1000 and ixgbe ... and appeared to work well on ixl. However, as it
turned out there was a race to this approach that would lockup the ixl MAC.
It was possible for a lower producer index to be written after a higher one.
On e1000 and ixgbe this was harmless - on ixl it was fatal. My initial
response was to add a lock around doorbell writes - fixing the problem but
adding an unacceptable amount of lock contention.
The next iteration was to use transmit interrupts to drive delayed doorbell
writes. If there were no packets in the queue all doorbell writes would be
immediate as the queue started to fill up we could delay doorbell writes
further and further. At the start of drain if we've cleaned any packets we
know we've moved the state machine along and we write the doorbell (an
obvious missing optimization was to skip that doorbell write if db_pending
is zero). This change required that tx interrupts be scheduled periodically
as opposed to just when the hardware txq was full. However, that just leads
to our next problem.
Initially dedicated msix vectors were used for both tx and rx. However, it
was often possible to use up all available vectors before we set up all the
queues we wanted. By having rx and tx share a vector for a given queue we
could halve the number of vectors used by a given configuration. The problem
here is that with this change only e1000 passed the necessary value to have
the fast interrupt drive tx when appropriate.
Reported by: mav@
Tested by: mav@
Reviewed by: gallatin@
MFC after: 1 month
Sponsored by: iXsystems
Differential Revision: https://reviews.freebsd.org/D27683
The pidx argument of isc_rxd_flush() indicates which is the last valid
receive descriptor to be used by the NIC. However, current code has
multiple issues:
- Intel drivers write pidx to their RDT register, which means that
NICs will only use the descriptors up to pidx-1 (modulo ring size N),
and won't actually use the one pointed by pidx. This does not break
reception, but it is anyway confusing and suboptimal (the NIC will
actually see only N-2 descriptors as available, rather than N-1).
Other drivers (if_vmx, if_bnxt, if_mgb) adhere to this semantic).
- The semantic used by Intel (RDT is one descriptor past the last
valid one) is used by most (if not all) NICs, and it is also used
on the TX side (also in iflib). Since iflib is not currently
using this semantic for RX, it must decrement fl->ifl_pidx
(modulo N) before calling isc_rxd_flush(), and then the
per-driver callback implementation must increment the index
again (to match the real semantic). This is confusing and suboptimal.
- The iflib refill function is also called at initialization.
However, in case the ring size is smaller than 128 (e.g. if_mgb),
the refill function will actually prepare all the receive
descriptors (N), without leaving one unused, as most of NICs assume
(e.g. to avoid RDT to overrun RDH). I can speculate that the code
looks like this right now because this issue showed up during
testing (e.g. with if_mgb), and it was easy to workaround by
decrementing pidx before isc_rxd_flush().
The goal of this change is to simplify the code (removing a bunch
of instructions from the RX fast path), and to make the semantic of
isc_rxd_flush() consistent across drivers. To achieve this, we:
- change the semantics of the pidx argument to the usual one (that
is the index one past the last valid one), so that both iflib and
drivers avoid the decrement/increment dance.
- fix the initialization code to prepare at most N-1 descriptors.
Reviewed by: markj
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D26191
Check copyin's error code (differ adding copyout checks at this time).
Don't directly access user memory in the switch statement.
Since bnxt_ioctl_data isn't all that big, use a stack allocation.
Reviewed by: jhb
MFC after: 3 days
Sponsored by: DARPA
Differential Revision: https://reviews.freebsd.org/D23933
r357614 added CTLFLAG_NEEDGIANT to make it easier to find nodes that are
still not MPSAFE (or already are but aren’t properly marked).
Use it in preparation for a general review of all nodes.
This is non-functional change that adds annotations to SYSCTL_NODE and
SYSCTL_PROC nodes using one of the soon-to-be-required flags.
Mark all obvious cases as MPSAFE. All entries that haven't been marked
as MPSAFE before are by default marked as NEEDGIANT
Approved by: kib (mentor, blanket)
Commented by: kib, gallatin, melifaro
Differential Revision: https://reviews.freebsd.org/D23718
While here, clean up magic numbers.
The memset(,0,) (and M_ZERO!) can just be removed; the bit_alloc() API already
zeros the allocation.
No functional change.
Reported by: Coverity
CID: 1378286
- Ever since the workaround for the silicon bug of TSO4 causing MAC hangs
was committed in r295133, CSUM_TSO always got disabled unconditionally
by em(4) on the first invocation of em_init_locked(). However, even with
that problem fixed, it turned out that for at least e. g. 82579 not all
necessary TSO workarounds are in place, still causing MAC hangs even at
Gigabit speed. Thus, for stable/11, TSO usage was deliberately disabled
in r323292 (r323293 for stable/10) for the EM-class by default, allowing
users to turn it on if it happens to work with their particular EM MAC
in a Gigabit-only environment.
In head, the TSO workaround for speeds other than Gigabit was lost with
the conversion to iflib(9) in r311849 (possibly along with another one
or two TSO workarounds). Yet at the same time, for EM-class MACs TSO4
got enabled by default again, causing device hangs. Therefore, change the
default for this hardware class back to have TSO4 off, allowing users
to turn it on manually if it happens to work in their environment as
we do in stable/{10,11}. An alternative would be to add a whitelist of
EM-class devices where TSO4 actually is reliable with the workarounds in
place, but given that the advantage of TSO at Gigabit speed is rather
limited - especially with the overhead of these workarounds -, that's
really not worth it. [1]
This change includes the addition of an isc_capabilities to struct
if_softc_ctx so iflib(9) can also handle interface capabilities that
shouldn't be enabled by default which is used to handle the default-off
capabilities of e1000 as suggested by shurd@ and moving their handling
from em_setup_interface() to em_if_attach_pre() accordingly.
- Although 82543 support TSO4 in theory, the former lem(4) didn't have
support for TSO4, presumably because TSO4 is even more broken in the
LEM-class of MACs than the later EM ones. Still, TSO4 for LEM-class
devices was enabled as part of the conversion to iflib(9) in r311849,
causing device hangs. So revert back to the pre-r311849 behavior of
not supporting TSO4 for LEM-class at all, which includes not creating
a TSO DMA tag in iflib(9) for devices not having IFCAP_TSO4 set. [2]
- In fact, the FreeBSD TCP stack can handle a TSO size of IP_MAXPACKET
(65535) rather than FREEBSD_TSO_SIZE_MAX (65518). However, the TSO
DMA must have a maxsize of the maximum TSO size plus the size of a
VLAN header for software VLAN tagging. The iflib(9) converted em(4),
thus, first correctly sets scctx->isc_tx_tso_size_max to EM_TSO_SIZE
in em_if_attach_pre(), but later on overrides it with IP_MAXPACKET
in em_setup_interface() (apparently, left-over from pre-iflib(9)
times). So remove the later and correct iflib(9) to correctly cap
the maximum TSO size reported to the stack at IP_MAXPACKET. While at
it, let iflib(9) use if_sethwtsomax*().
This change includes the addition of isc_tso_max{seg,}size DMA engine
constraints for the TSO DMA tag to struct if_shared_ctx and letting
iflib_txsd_alloc() automatically adjust the maxsize of that tag in case
IFCAP_VLAN_MTU is supported as requested by shurd@.
- Move the if_setifheaderlen(9) call for adjusting the maximum Ethernet
header length from {ixgbe,ixl,ixlv,ixv,em}_setup_interface() to iflib(9)
so adjustment is automatically done in case IFCAP_VLAN_MTU is supported.
As a consequence, this adjustment now is also done in case of bnxt(4)
which missed it previously.
- Move the reduction of the maximum TSO segment count reported to the
stack by the number of m_pullup(9) calls (which in the worst case,
can add another mbuf and, thus, the requirement for another DMA
segment each) in the transmit path for performance reasons from
em_setup_interface() to iflib_txsd_alloc() as these pull-ups are now
done in iflib_parse_header() rather than in the no longer existing
em_xmit(). Moreover, this optimization applies to all drivers using
iflib(9) and not just em(4); all in-tree iflib(9) consumers still
have enough room to handle full size TSO packets. Also, reduce the
adjustment to the maximum number of m_pullup(9)'s now performed in
iflib_parse_header().
- Prior to the conversion of em(4)/igb(4)/lem(4) and ixl(4) to iflib(9)
in r311849 and r335338 respectively, these drivers didn't enable
IFCAP_VLAN_HWFILTER by default due to VLAN events not being passed
through by lagg(4). With iflib(9), IFCAP_VLAN_HWFILTER was turned on
by default but also lagg(4) was fixed in that regard in r203548. So
just remove the now redundant and defunct IFCAP_VLAN_HWFILTER handling
in {em,ixl,ixlv}_setup_interface().
- Nuke other redundant IFCAP_* setting in {em,ixl,ixlv}_setup_interface()
which is (more completely) already done in {em,ixl,ixlv}_if_attach_pre()
now.
- Remove some redundant/dead setting of scctx->isc_tx_csum_flags in
em_if_attach_pre().
- Remove some IFCAP_* duplicated either directly or indirectly (e. g.
via IFCAP_HWCSUM) in {EM,IGB,IXL}_CAPS.
- Don't bother to fiddle with IFCAP_HWSTATS in ixgbe(4)/ixgbev(4) as
iflib(9) adds that capability unconditionally.
- Remove some unused macros from em(4).
- Bump __FreeBSD_version as some of the above changes require the modules
of drivers using iflib(9) to be recompiled.
Okayed by: sbruno@ at 201806 DevSummit Transport Working Group [1]
Reviewed by: sbruno (earlier version), erj
PR: 219428 (part of; comment #10) [1], 220997 (part of; comment #3) [2]
Differential Revision: https://reviews.freebsd.org/D15720
Uses of mallocarray(9).
The use of mallocarray(9) has rocketed the required swap to build FreeBSD.
This is likely caused by the allocation size attributes which put extra pressure
on the compiler.
Given that most of these checks are superfluous we have to choose better
where to use mallocarray(9). We still have more uses of mallocarray(9) but
hopefully this is enough to bring swap usage to a reasonable level.
Reported by: wosch
PR: 225197
Focus on code where we are doing multiplications within malloc(9). None of
these is likely to overflow, however the change is still useful as some
static checkers can benefit from the allocation attributes we use for
mallocarray.
This initial sweep only covers malloc(9) calls with M_NOWAIT. No good
reason but I started doing the changes before r327796 and at that time it
was convenient to make sure the sorrounding code could handle NULL values.
Previously, it silently only supported auto, instead, log a message
indicating why only auto is supported.
Submitted by: Bhargava Chenna Marreddy <bhargava.marreddy@broadcom.com>
Sponsored by: Broadcom Limited
Differential Revision: https://reviews.freebsd.org/D13358
It's possible to change the SFP module when link is down, which would
change the available media types. This is part of D13358.
Submitted by: Bhargava Chenna Marreddy <bhargava.marreddy@broadcom.com>
Sponsored by: Broadcom Limited
New Stratus bnxt devices require support for short HWRM commands for VFs
to function. Enable their use, but only use them if it's both supported
and required... prefer the long HWRM commands when possible.
Submitted by: Bhargava Chenna Marreddy <bhargava.marreddy@broadcom.com>
Sponsored by: Broadcom Limited
Differential Revision: https://reviews.freebsd.org/D13269?id=36180
The firmware is always in little endian, use htole*() for all request fields
larger than one byte.
Submitted by: Bhargava Chenna Marreddy <bhargava.marreddy@broadcom.com>
Sponsored by: Broadcom Limited
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
Use our ifm_list of supported media types rather than nested switch
statements to find the current media type. Find a supported type that
matches the current speed.
Remove all workarounds while updating ifmr->ifm_active.
For BNXT_IFMEDIA_ADD, added Three more speeds IFM_10G_T, IFM_2500_T & IFM_2500_KX.
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/D12896
1) Add new phy_types and speeds from the latest firmware header.
2) Introduced a macro to avoid code duplication and improve readability for
the invocation of ifmedia_add().
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/D12423
Improved logging added in r323879 exposed an error during
attach. We need the irq, not the rid to work correctly. em uses
shared irqs, so it will use the same irq for TX as RX. bnxt does
not use shared irqs, or TX irqs at all, so there's no need to set
the TX irq affinity.
Reviewed by: sbruno
Approved by: sbruno (mentor)
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D12496
GPUs: radeonkms, i915kms
NICs: if_em, if_igb, if_bnxt
This metadata isn't used yet, but it will be handy to have later to
implement automatic module loading.
Reviewed by: imp, mmacy
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D12488
1) Choose correct Firmware options for HW LRO for best performance
2) Delete TBD and other comments which are not required.
3) Added sysctl interface to enable / disable / modify different factors
of HW LRO.
4) Disabled HW LRO by default to avoid issues with packet forwarding
This allows much better control over the LRO configuration via sysctls, and
uses much better defaults. Hardware LRO can now be enabled/disabled
independantly from the software LRO, and the tuning parameters are exposed.
manpage updates coming soon.
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/D12223
- Use HWRM_FUNC_VF_CFG instead of HWRM_FUNC_CFG on VFs
- Fix NPAR/VF detection
- Clean up flag definitions
- Don't allow WoL on VFs
Although the bnxt driver doesn't support SR-IOV so can create VFs yet,
the PF could be running Linux or ESCi with a VF passed through to a
FreeBSD guest. This fixes the driver for that use case.
Submitted by: Siva Kallam <siva.kallam@@broadcom.com>
Reviewed by: shurd, sbruno
Approved by: sbruno (mentor)
Sponsored by: Broadcom Limited
Differential Revision: https://reviews.freebsd.org/D12410
This was really too big of a commit even if everything worked, but there
are multiple new issues introduced in the one huge commit, so it's not
worth keeping this until it's fixed.
I'll work on splitting this up into logical chunks and introduce them one
at a time over the next week or two.
Approved by: sbruno (mentor)
Sponsored by: Limelight Networks
by Matt Macy as well as other changes which he has accepted via pull
request to his github repo at https://github.com/mattmacy/networking/
This should bring -CURRENT and the github repo into close enough sync to
allow small feature branches rather than a large chain of interdependant
patches being developed out of tree. The reset of the synchronization
should be able to be completed on github by splitting the remaining
changes that are not yet ready into short feature branches for later
review as smaller commits.
Here is a summary of changes included in this patch:
1) More checks when INVARIANTS are enabled for eariler problem
detection
2) Group Task Queue cleanups
- Fix use of duplicate shortdesc for gtaskqueue malloc type.
Some interfaces such as memguard(9) use the short description to
identify malloc types, so duplicates should be avoided.
3) Allow gtaskqueues to use ithreads in addition to taskqueues
- In some cases, this can improve performance
4) Better logging when taskqgroup_attach*() fails to set interrupt
affinity.
5) Do not start gtaskqueues until they're needed
6) Have mp_ring enqueue function enter the ABDICATED rather than BUSY
state. This moves the TX to the gtaskq and allows processing to
continue faster as well as make TX batching more likely.
7) Add an ift_txd_errata function to struct if_txrx. This allows
drivers to inspect/modify mbufs before transmission.
8) Add a new IFLIB_NEED_ZERO_CSUM for drivers to indicate they need
checksums zeroed for checksum offload to work. This avoids modifying
packet data in the TX path when possible.
9) Use ithreads for iflib I/O instead of taskqueues
10) Clean up ioctl and support async ioctl functions
11) Prefetch two cachlines from each mbuf instead of one up to 128B. We
often need to parse packet header info beyond 64B.
12) Fix potential memory corruption due to fence post error in
bit_nclear() usage.
13) Improved hang detection and handling
14) If the packet is smaller than MTU, disable the TSO flags.
This avoids extra packet parsing when not needed.
15) Move TCP header parsing inside the IS_TSO?() test.
This avoids extra packet parsing when not needed.
16) Pass chains of mbufs that are not consumed by lro to if_input()
rather call if_input() for each mbuf.
17) Re-arrange packet header loads to get as much work as possible done
before a cache stall.
18) Lock the context when calling IFDI_ATTACH_PRE()/IFDI_ATTACH_POST()/
IFDI_DETACH();
19) Attempt to distribute RX/TX tasks across cores more sensibly,
especially when RX and TX share an interrupt. RX will attempt to
take the first threads on a core, and TX will attempt to take
successive threads.
20) Allow iflib_softirq_alloc_generic() to request affinity to the same
cpus an interrupt has affinity with. This allows TX queues to
ensure they are serviced by the socket the device is on.
21) Add new iflib sysctls to net.iflib:
- timer_int - interval at which to run per-queue timers in ticks
- force_busdma
22) Add new per-device iflib sysctls to dev.X.Y.iflib
- rx_budget allows tuning the batch size on the RX path
- watchdog_events Count of watchdog events seen since load
23) Fix error where netmap_rxq_init() could get called before
IFDI_INIT()
24) e1000: Fixed version of r323008: post-cold sleep instead of DELAY
when waiting for firmware
- After interrupts are enabled, convert all waits to sleeps
- Eliminates e1000 software/firmware synchronization busy waits after
startup
25) e1000: Remove special case for budget=1 in em_txrx.c
- Premature optimization which may actually be incorrect with
multi-segment packets
26) e1000: Split out TX interrupt rather than share an interrupt for
RX and TX.
- Allows better performance by keeping RX and TX paths separate
27) e1000: Separate igb from em code where suitable
Much easier to understand separate functions and "if (is_igb)" than
previous tests like "if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))"
#blamebruno
Reviewed by: sbruno
Approved by: sbruno (mentor)
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D12235
This provides port stats (updated once per second) in
dev.bnxt.X.port_stats for PFs. VFs do not have access to the port stats.
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/D11914
hsi_struct_def.h file contains all firmware (HWRM) data struct's, updated
that with the latest one which was released on 30'th Aug.
After this upgrade, HWRM version will be 1.8.1.5 (earlier it was 1.4.0).
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/D12203
1) Based on the suggestion from firmware team, derive
scctx->isc_ntxqsets_max & scctx->isc_nrxqsets_max based on FUNC_QCFG
(instead of FUNC_QCAPS).
2) Bump-up driver version to "1.0.0.2".
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/D12128
were redundant and not being used to set anything up.
Submitted by: Matt Macy <mmacy@mattmacy.io>
Reported by: Jeb Cramer <cramerj@intel.com>
Sponsored by: Limelight Networks
iflib - reset fl-ifl_fragidx to 0 on iflib_fl_bufs_free(). This caused the
panic in em/igb when adding it to a bridge device.
iflib - Handle out of order packet delivery from hardware in support of LRO
Out of order updates to rxd's is fixed in r315217. However, it is not
completely fixed. While refilling the buffers, iflib is not considering
the out of order descriptors. Hence, it is refilling sequentially.
"idx" variable in _iflib_fl_refill routine is incremented sequentially.
By doing refilling sequentially, it will override the SGEs that
are *IN USE* by other connections. Fix is to maintain a bitmap of
rx descriptors and differentiate the used one with unused one and
refill only at the unused indices. This patch also fixes a
few bugs in bnxt, related to the same feature.
Submitted by: bhargava.marreddy@broadcom.com
Reviewed by: venkatkumar.duvvuru@broadcom.com shurd
Differential Revision: https://reviews.freebsd.org/D10681
--Remove special-case handling of sparc64 bus_dmamap* functions.
Replace with a more generic mechanism that allows MD busdma
implementations to generate inline mapping functions by
defining WANT_INLINE_DMAMAP in <machine/bus_dma.h>. This
is currently useful for sparc64, x86, and arm64, which all
implement non-load dmamap operations as simple wrappers
around map objects which may be bus- or device-specific.
--Remove NULL-checked bus_dmamap macros. Implement the
equivalent NULL checks in the inlined x86 implementation.
For non-x86 platforms, these checks are a minor pessimization
as those platforms do not currently allow NULL maps. NULL
maps were originally allowed on arm64, which appears to have
been the motivation behind adding arm[64]-specific barriers
to bus_dma.h, but that support was removed in r299463.
--Simplify the internal interface used by the bus_dmamap_load*
variants and move it to bus_dma_internal.h
--Fix some drivers that directly include sys/bus_dma.h
despite the recommendations of bus_dma(9)
Reviewed by: kib (previous revision), marius
Differential Revision: https://reviews.freebsd.org/D10729
This generates startup LORs and panics when adding elements to bridge
devices. I will document further in https://reviews.freebsd.org/D10681
PR: 220073
Submitted by: dchagin
Reported by: db
iflib - Handle out of order packet delivery from hardware in support of LRO
Out of order updates to rxd's is fixed in r315217. However, it is not
completely fixed. While refilling the buffers, iflib is not considering
the out of order descriptors. Hence, it is refilling sequentially.
"idx" variable in _iflib_fl_refill routine is incremented sequentially.
By doing refilling sequentially, it will override the SGEs that
are *IN USE* by other connections. Fix is to maintain a bitmap of
rx descriptors and differentiate the used one with unused one and
refill only at the unused indices. This patch also fixes a
few bugs in bnxt, related to the same feature.
Submitted by: bhargava.marreddy@broadcom.com
Reviewed by: shurd@
Differential Revision: https://reviews.freebsd.org/D10681
- unconditionally enable BUS_DMA on non-x86 architectures
- speed up rxd zeroing via customized function
- support out of order updates to rxd's
- add prefetching to hardware descriptor rings
- only prefetch on 10G or faster hardware
- add seperate tx queue intr function
- preliminary rework of NETMAP interfaces, WIP
Submitted by: Matt Macy <mmacy@nextbsd.org>
Sponsored by: Limelight Networks
