em/emx: Sync with Intel's em 7.6.2

- Fix RX filter settings when exiting promiscous mode.
- Add workaround for I219 resetting.
- Fix some settings for 82574.

Tested-by: dillon@ (I219)

emx: Free tx mbufs proactively.

This is preparation for the dillon's upcoming sendfile patch.

emx: Use ringmap to do polling cpu assignment and fill redirect table.

em/emx: Integrate ifmedia flow control support.

Other minor bits:
- Set media to IFM_NONE, if there is no link yet.
- Remove no longer used dev.em.Y.flow_ctrl sysctl node.
ifconfig emY [mediaopt|-

em/emx: Integrate ifmedia flow control support.

Other minor bits:
- Set media to IFM_NONE, if there is no link yet.
- Remove no longer used dev.em.Y.flow_ctrl sysctl node.
ifconfig emY [mediaopt|-mediaopt] [rxpause|txpause|forcepause]
should be used instead.

show more ...

emx: Support flow control change and default to RX pause

Tested-by: dillon@

kernel: Use the new auto-created sysctl ctx/tree in various drivers.

Everywhere where we formerly rolled a hw.fooX.* per-device ctx/tree
ourselves.

This commit switches it to dev.foo.X.*

emx: Add I217/I218

According to the datasheet, these two chips support 2 RX queues at least.

em/emx: Avoid using magic number for TX register writing aggregation

emx: Implement multiple hardware TX queue support; disabled by default

Two notes about the multiple hardware TX queue support:

- It could only be enabled on 82571, 82572 and es2lan. Enabling it on

emx: Implement multiple hardware TX queue support; disabled by default

Two notes about the multiple hardware TX queue support:

- It could only be enabled on 82571, 82572 and es2lan. Enabling it on
82574, whose datasheet claims multiple hardware TX queue support,
could cause watchdog timeout when heavy TCP receiving test are
conducted.

- It is disabled by default. Since 82571, 82572 and es2lan only have
one global context for TX hardware checksum and TSO, the driver will
have to setup TX context descriptor for each packet. Using TX context
descriptor for each packet could cause serious transmission performance
regression for tiny packets, see comment of following commit:
51e6819fc2ad7cefdf830dcc9f20bcbe2af2aa52

However, multiple hardware TX queues still could be enabled depending
on the workload (*):
o TSO probably are not affected by setting up a TX context descriptor
for each TSO packet.
o Stuffs do not using hardware TX checksum, e.g. IP fast forwarding.

Multiple hardware TX queues could be enabled globally by tunable:
hw.emx.txr=2
Multiple hardware TX queues could be enabled on individual interface by
tunable:
hw.emxY.txr=2

(*) Another thought is to add per-interface metric, which indicates the
size of packets that worth hardware checksum; if the packets' size is
below that metric, software checksum is used.

show more ...

emx: Don't update TDT upon very packets transmitted

The TDT is updated only after certain amount of TX descriptors are
added to the hardware TX ring. The default value of the amount of
TX descripto

emx: Don't update TDT upon very packets transmitted

The TDT is updated only after certain amount of TX descriptors are
added to the hardware TX ring. The default value of the amount of
TX descriptors are 8. This value could be further tuned by per-device
sysctl node hw.emxX.tx_wreg_nsegs.

CPU usage when transmitting 18bytes tiny UDP datagrams @1.48Mpps with
82571 on AMD970 (3500Mhz) are reduced from 50% to 30% with the default
value.

Inspired-by: FreeBSD netmap's paper

show more ...

emx: Group TX related fields into emx_txdata; use it in TX related functions

emx: Pass emx_rxdata to RX related functions

emx: Remove unused statistics

emx: Allow user to specify RX/TX processing CPU's offset

emx: Pack boolean fields into flags

emx: Increase spare TX descriprors to 33 incorporate TSO segment

em/emx: Reduce the default interrupt rate to 6000Hz

emx: Add EMX_TSO_DEBUG

It is used to measure TSO context reuse rate. Setting up TX context
interferes TX data fetching pipelining, thus reduces performance, so
TX context setting up should be avoid

emx: Add EMX_TSO_DEBUG

It is used to measure TSO context reuse rate. Setting up TX context
interferes TX data fetching pipelining, thus reduces performance, so
TX context setting up should be avoided as much as possible.

According to limited measurement:
- 1 one direction TX stream, reuse rate is ~40%
- 16 one direction TX streams, reuse rate is ~20%
- 16 bi-direction streams (total 32), reuse rate is ~10%

show more ...

emx: Add TSO support

emx: Remove unused macros

emx: Utilize mbuf's header lengthes

emx: Fix "Missing Interrupt Following ICR read" errata

IMS should be set to 0 before reading ICR. It should be noted that
once IMS is 0, ICR.INT_ASSERTED will not be set, which is not mentioned
in

emx: Fix "Missing Interrupt Following ICR read" errata

IMS should be set to 0 before reading ICR. It should be noted that
once IMS is 0, ICR.INT_ASSERTED will not be set, which is not mentioned
in the datasheet.

This errata exists on 82571, 82572, 82573, 82574 and 82583, while 82583
is not covered by emx(4). And the fix is needed only if the legacy
interrupt is used and the interrupt line is shared with other devices.

hw.emxX.irq.unshared tunable is added to give a hint to the driver that
the legacy interrupt is not shared with other devices.

show more ...

emx: Change how IFCAP_RSS is handled

- Number of RX rings being used is not changed.
- Hardware is always setup w/ RSS support.
- If IFCAP_RSS is not enabled, we simply don't use the RSS hash provid

emx: Change how IFCAP_RSS is handled

- Number of RX rings being used is not changed.
- Hardware is always setup w/ RSS support.
- If IFCAP_RSS is not enabled, we simply don't use the RSS hash provided
by the hardware.

show more ...

pci: Factor out pci_alloc_1intr()

This function will properly setup rid and flags for "legacy" interrupt.
If MSI is enabled, then this function will try allocate 1 MSI message
and setup rid and flag

pci: Factor out pci_alloc_1intr()

This function will properly setup rid and flags for "legacy" interrupt.
If MSI is enabled, then this function will try allocate 1 MSI message
and setup rid and flags properly for MSI based interrupt resource.

show more ...

emx: Use MSI, if device supports it