CVS log for src/sys/dev/pci/if_wm.c

Up to [cvs.NetBSD.org] / src / sys / dev / pci

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

Add MBUFTRACE

sys: Drop redundant NULL check before m_freem(9)

m_freem(9) safely has accepted NULL argument at least since 4.2BSD:
https://www.tuhs.org/cgi-bin/utree.pl?file=4.2BSD/usr/src/sys/sys/uipc_mbuf.c

Compile-tested on amd64/ALL.

Suggested by knakahara@

if_stats(9): Add ifp argument to if_stat..._ref.

This will enable us to pass the ifp through to a dtrace probe inside.

No functional change intended in this change, but this is an API
change visible to modules so it shouldn't be pulled up.

PR kern/58377

Pull up the following, requested by msaitoh in ticket #1938:

	sys/dev/pci/if_wm.c			1.792,1.794-1.798 via patch
	sys/dev/pci/if_wmreg.h			1.131

- Add RQDPC(Receive Queue Drop Packet Count) to iqdrops.
- Drop frames if the RX descriptor ring has no room on multiqueue system.
- Improve dmesg output.
  - Print RX packet buffer size.
  - Fix the upper 16bit of Image Unique ID(EtrackID).
- Fix comment.

Pull up the following, requested by msaitoh in ticket #1802:

	sys/dev/pci/if_wm.c			1.794-1.798 via patch
	sys/dev/pci/if_wmreg.h			1.131

- Add RQDPC(Receive Queue Drop Packet Count) to iqdrops.
- Drop frames if the RX descriptor ring has no room on multiqueue
  system.
- Improve dmesg output:
  - Print RX packet buffer size.
  - Fix the upper 16bit of Image Unique ID(EtrackID).

Pull up following revision(s) (requested by msaitoh in ticket #607):

	sys/dev/pci/if_wm.c: revision 1.795
	sys/dev/pci/if_wm.c: revision 1.796
	sys/dev/pci/if_wm.c: revision 1.797
	sys/dev/pci/if_wmreg.h: revision 1.131

wm(4): Add RQDPC(Receive Queue Drop Packet Count) to iqdrops.

 The iqdrops counter should include not only MPC(Missed Packet Count)
but also RQDPC(Receive Queue Drop Packet Count). Same as ixgbe(4) and igc(4).

 Note that the RQDPC is not currently counted because SRRCTL_DROP_EN
is not set.

wm(4): Drop frames if the RX descriptor ring has no room on multiqueue system.

 Drop frames if the RX descriptor ring has no room. This is enabled only on
multiqueue system to avoid bad influence to other queues. The drop count
can be seen by the RQDPC counter (wmN rxqXXdrop evcnt).

wm(4): Fix compile error without WM_EVENT_COUNTERS.

Pull up following revision(s) (requested by msaitoh in ticket #606):

	sys/dev/pci/if_wm.c: revision 1.794
	sys/dev/pci/if_wm.c: revision 1.798

wm(4): Print RX packet buffer size.

wm(4): Fix upper 16bit of Image Unique ID(EtrackID).

 Don't override uid1 variable while reading option ROM version
to print Image Unique ID correctly.

wm(4): Fix upper 16bit of Image Unique ID(EtrackID).

 Don't override uid1 variable while reading option ROM version
to print Image Unique ID correctly.

Pull up following revision(s) (requested by msaitoh in ticket #1932):

	sys/dev/pci/if_wm.c: revision 1.793

Add Intel I219-{LM,V}(20,21) support.

Pull up following revision(s) (requested by msaitoh in ticket #1791):

	sys/dev/pci/if_wm.c: revision 1.793

Add Intel I219-{LM,V}(20,21) support.

Pull up following revision(s) (requested by msaitoh in ticket #564):

	sys/dev/pci/if_wm.c: revision 1.793

Add Intel I219-{LM,V}(20,21) support.

wm(4): Fix compile error without WM_EVENT_COUNTERS.

wm(4): Drop frames if the RX descriptor ring has no room on multiqueue system.

 Drop frames if the RX descriptor ring has no room. This is enabled only on
multiqueue system to avoid bad influence to other queues. The drop count
can be seen by the RQDPC counter (wmN rxqXXdrop evcnt).

wm(4): Add RQDPC(Receive Queue Drop Packet Count) to iqdrops.

 The iqdrops counter should include not only MPC(Missed Packet Count)
but also RQDPC(Receive Queue Drop Packet Count). Same as ixgbe(4) and igc(4).

 Note that the RQDPC is not currently counted because SRRCTL_DROP_EN
is not set.

wm(4): Print RX packet buffer size.

Add Intel I219-{LM,V}(20,21) support.

if_wm.c: fix spelling and grammar in a comment

Pull up the following, requested by msaitoh in ticket #1915:

	sys/dev/pci/pcidevs			1.1497
	sys/dev/pci/if_wm.c			1.689,1.790-1.791 via patch
	sys/dev/pci/if_wmreg.h			1.120 via patch
	sys/dev/pci/if_wmvar.h			1.51

- Use 12K for packet buffer for jumbo frame on PCH2 and newer.
- Add new workaround for Tiger Lake and newer to avoid packet loss.
- Add I219{V,LM}({22,23}) devices (Raptor Lake).

Pull up following revision(s) (requested by msaitoh in ticket #1754):

	sys/dev/pci/pcidevs: revision 1.1497
	sys/dev/pci/if_wm.c: revision 1.790
	sys/dev/pci/if_wm.c: revision 1.791
	sys/dev/pci/if_wmvar.h: revision 1.51

pcidevs: Add Intel I219{V,LM}({22,23})

wm(4): Add a new workaround for Tiger Lake and newer.
 - Define new WM_T_PCH_TGP and use it for Tiger Lake and newer.
   Note that we don't define WM_T_PCH_ADP because we have no any
   Alder Lake specific workaround yet.
 - Add new workaround for Tiger Lake (and newer) in wm_init_locked()
   to avoid packet loss.

wm(4): Add I219{V,LM}({22,23}) devices (Raptor Lake).

Pull up following revision(s) (requested by msaitoh in ticket #421):

	sys/dev/pci/pcidevs: revision 1.1497
	sys/dev/pci/if_wm.c: revision 1.790
	sys/dev/pci/if_wm.c: revision 1.791
	sys/dev/pci/if_wmvar.h: revision 1.51

pcidevs: Add Intel I219{V,LM}({22,23})

wm(4): Add a new workaround for Tiger Lake and newer.
 - Define new WM_T_PCH_TGP and use it for Tiger Lake and newer.
   Note that we don't define WM_T_PCH_ADP because we have no any
   Alder Lake specific workaround yet.
 - Add new workaround for Tiger Lake (and newer) in wm_init_locked()
   to avoid packet loss.

wm(4): Add I219{V,LM}({22,23}) devices (Raptor Lake).

wm(4): Add I219{V,LM}({22,23}) devices (Raptor Lake).

wm(4): Add a new workaround for Tiger Lake and newer.

 - Define new WM_T_PCH_TGP and use it for Tiger Lake and newer.
   Note that we don't define WM_T_PCH_ADP because we have no any
   Alder Lake specific workaround yet.
 - Add new workaround for Tiger Lake (and newer) in wm_init_locked()
   to avoid packet loss.

Pull up following revision(s) (requested by msaitoh in ticket #1910):

	sys/dev/pci/if_wm.c: revision 1.785
	sys/dev/pci/if_wm.c: revision 1.786
	sys/dev/pci/if_wm.c: revision 1.787
	sys/dev/pci/if_wm.c: revision 1.788
	sys/dev/pci/if_wm.c: revision 1.789

wm(4): Use SCVPC and HRMPC for 82575 and newer.
  The SCVPC(SerDes/SGMII Code Violation Packet Count) register and the HRMPC
  (Header Redirection Missed Packet Count) register were added in if_wm.c
  rev. 1.776 but the location in the code were incorrect. Fix them.

wm(4): Add some info to some event counters.
 - The difference between the RUC(Rx Undersize) and RFC(Rx Fragment) is whether
   the CRC is valid or not.
 - The difference between the ROC(Rx Oversize) and RJC(Rx Jabber) is whether
   the CRC is valid or not.
 - LENERRS(Length Errors) is for the length/type field <= 1500.

wm(4): the IAC(Interrupt Assertion Count) register is for 82571 and newer.

wm(4): Modify descriptions of flow control related event counters.

wm(4): Replace /* nothing */ to __nothing for evcnt macros
Prevent empty if bodies for !WM_EVENT_COUNTERS.

Pull up following revision(s) (requested by msaitoh in ticket #1746):

	sys/dev/pci/if_wm.c: revision 1.785
	sys/dev/pci/if_wm.c: revision 1.786
	sys/dev/pci/if_wm.c: revision 1.787
	sys/dev/pci/if_wm.c: revision 1.788
	sys/dev/pci/if_wm.c: revision 1.789

wm(4): Use SCVPC and HRMPC for 82575 and newer.
  The SCVPC(SerDes/SGMII Code Violation Packet Count) register and the HRMPC
  (Header Redirection Missed Packet Count) register were added in if_wm.c
  rev. 1.776 but the location in the code were incorrect. Fix them.

wm(4): Add some info to some event counters.
 - The difference between the RUC(Rx Undersize) and RFC(Rx Fragment) is whether
   the CRC is valid or not.
 - The difference between the ROC(Rx Oversize) and RJC(Rx Jabber) is whether
   the CRC is valid or not.
 - LENERRS(Length Errors) is for the length/type field <= 1500.

wm(4): the IAC(Interrupt Assertion Count) register is for 82571 and newer.

wm(4): Modify descriptions of flow control related event counters.

wm(4): Replace /* nothing */ to __nothing for evcnt macros
Prevent empty if bodies for !WM_EVENT_COUNTERS.

Pull up following revision(s) (requested by msaitoh in ticket #396):

	sys/dev/pci/if_wm.c: revision 1.785
	sys/dev/pci/if_wm.c: revision 1.786
	sys/dev/pci/if_wm.c: revision 1.787
	sys/dev/pci/if_wm.c: revision 1.788
	sys/dev/pci/if_wm.c: revision 1.789

wm(4): Use SCVPC and HRMPC for 82575 and newer.
  The SCVPC(SerDes/SGMII Code Violation Packet Count) register and the HRMPC
  (Header Redirection Missed Packet Count) register were added in if_wm.c
  rev. 1.776 but the location in the code were incorrect. Fix them.

wm(4): Add some info to some event counters.
 - The difference between the RUC(Rx Undersize) and RFC(Rx Fragment) is whether
   the CRC is valid or not.
 - The difference between the ROC(Rx Oversize) and RJC(Rx Jabber) is whether
   the CRC is valid or not.
 - LENERRS(Length Errors) is for the length/type field <= 1500.

wm(4): the IAC(Interrupt Assertion Count) register is for 82571 and newer.

wm(4): Modify descriptions of flow control related event counters.

wm(4): Replace /* nothing */ to __nothing for evcnt macros
Prevent empty if bodies for !WM_EVENT_COUNTERS.

wm(4): Replace /* nothing */ to __nothing for evcnt macros

Prevent empty if bodies for !WM_EVENT_COUNTERS.

wm(4): Modify descriptions of flow control related event counters.

wm(4): the IAC(Interrupt Assertion Count) register is for 82571 and newer.

wm(4): Add some info to some event counters.

 - The difference between the RUC(Rx Undersize) and RFC(Rx Fragment) is whether
   the CRC is valid or not.
 - The difference between the ROC(Rx Oversize) and RJC(Rx Jabber) is whether
   the CRC is valid or not.
 - LENERRS(Length Errors) is for the length/type field <= 1500.

wm(4): Use SCVPC and HRMPC for 82575 and newer.

  The SCVPC(SerDes/SGMII Code Violation Packet Count) register and the HRMPC
  (Header Redirection Missed Packet Count) register were added in if_wm.c
  rev. 1.776 but the location in the code were incorrect. Fix them.

Pull up following revision(s) (requested by msaitoh in ticket #1894):

	sys/dev/pci/if_wmvar.h: revision 1.50
	sys/dev/pci/if_wm.c: revision 1.783,1.784 via patch

Delay sending LINK_STATE_UP to prevent dropping packets on I35[04] and I21[01].

 Some (not all) systems use I35[04] or I21[01] don't send packet soon
after linkup. The MAC send a packet to the PHY and any error is not
observed. This behavior causes a problem that gratuitous ARP and/or
IPv6 DAD packet are silently dropped. To avoid this problem, don't
call mii_pollstat() here which will send LINK_STATE_UP notification
to the upper layer. Instead, mii_pollstat() will be called in
wm_gmii_mediastatus() or mii_tick() will be called in wm_tick().

Note that the similar workaround is in Linux's igb driver though it's
only for I21[01].

OK'd by hikaru@ and knakahara@.

Fix #ifdef WM_DEBUG code in wm_gmii_i82544_{read,write}reg_locked.

Pull up following revision(s) (requested by msaitoh in ticket #344):

	sys/dev/pci/if_wmvar.h: revision 1.50
	sys/dev/pci/if_wm.c: revision 1.783,1.784 via patch

Delay sending LINK_STATE_UP to prevent dropping packets on I35[04] and I21[01].

 Some (not all) systems use I35[04] or I21[01] don't send packet soon
after linkup. The MAC send a packet to the PHY and any error is not
observed. This behavior causes a problem that gratuitous ARP and/or
IPv6 DAD packet are silently dropped. To avoid this problem, don't
call mii_pollstat() here which will send LINK_STATE_UP notification
to the upper layer. Instead, mii_pollstat() will be called in
wm_gmii_mediastatus() or mii_tick() will be called in wm_tick().

Note that the similar workaround is in Linux's igb driver though it's
only for I21[01].

OK'd by hikaru@ and knakahara@.

Fix #ifdef WM_DEBUG code in wm_gmii_i82544_{read,write}reg_locked.

Pull up following revision(s) (requested by msaitoh in ticket #344):

	sys/dev/pci/if_wmvar.h: revision 1.50
	sys/dev/pci/if_wm.c: revision 1.783
	sys/dev/pci/if_wm.c: revision 1.784

Delay sending LINK_STATE_UP to prevent dropping packets on I35[04] and I21[01].

 Some (not all) systems use I35[04] or I21[01] don't send packet soon
after linkup. The MAC send a packet to the PHY and any error is not
observed. This behavior causes a problem that gratuitous ARP and/or
IPv6 DAD packet are silently dropped. To avoid this problem, don't
call mii_pollstat() here which will send LINK_STATE_UP notification
to the upper layer. Instead, mii_pollstat() will be called in
wm_gmii_mediastatus() or mii_tick() will be called in wm_tick().

Note that the similar workaround is in Linux's igb driver though it's
only for I21[01].

OK'd by hikaru@ and knakahara@.

Fix #ifdef WM_DEBUG code in wm_gmii_i82544_{read,write}reg_locked.

Fix #ifdef WM_DEBUG code in wm_gmii_i82544_{read,write}reg_locked.

Delay sending LINK_STATE_UP to prevent dropping packets on I35[04] and I21[01].

 Some (not all) systems use I35[04] or I21[01] don't send packet soon
after linkup. The MAC send a packet to the PHY and any error is not
observed. This behavior causes a problem that gratuitous ARP and/or
IPv6 DAD packet are silently dropped. To avoid this problem, don't
call mii_pollstat() here which will send LINK_STATE_UP notification
to the upper layer. Instead, mii_pollstat() will be called in
wm_gmii_mediastatus() or mii_tick() will be called in wm_tick().

Note that the similar workaround is in Linux's igb driver though it's
only for I21[01].

OK'd by hikaru@ and knakahara@.

Pull up the following revisions, requested by msaitoh in #1847:

	sys/dev/pci/if_wm.c				1.768-1.782 via patch
	sys/dev/pci/if_wmreg.h				1.129-1.130
	sys/dev/pci/if_wmvar.h				1.49

wm(4):
- Rework for event counters:
  - Fix calculation of GORC, GOTC, TOR and TOT counters correctly.
  - Rearrange the order of the registers so that they are roughly
    in ascending order.
  - Reorder evcnt_attach_dynamic(), WM_EVCNT_ADD() and evcnt_detach()
    to match.
  - IC{TX,RX}*C registers are for older than 82575.
  - Fix a bug that the transmit underrun counter is incorrectly
    counted.
  - Don't add "Count" for event counter's description.
  - Some statistics registers were replaced with new counters on newer
    chips. Treat 0x403c(CEXTERR->HTDPMC), 0x40fc(TSCTFC->CBRMPC),
    0x4124(ICRXOC->HTCBDPC) and from 0x4104 to 0x4124.
  - Add some new counters:
    - Circuit Breaker TX Manageability Packet
    - Circuit Breaker RX Dropped Packet
    - Host Good Octets RX
    - Host Good Octets TX
    - Length Errors
    - SerDes/SGMII Code Violation Packet
    - Header Redirection Missed Packet
    - EEE TX LPI
    - EEE RX LPI
  - Fix prc511's comment and description.
- Add SOICZIFDATA (ifconfig -z) support for evcnt(9).
- Use WM_IS_ICHPCH(). No functional change.
- Fix typo. s/ictxact/ictxatc/. No functional change.
- Add comment.

Pull up the following revisions, requested by msaitoh in #1656:

	sys/dev/pci/if_wm.c				1.768-1.782 via patch
	sys/dev/pci/if_wmreg.h				1.129-1.130
	sys/dev/pci/if_wmvar.h				1.49

wm(4):
- Rework for event counters:
  - Fix calculation of GORC, GOTC, TOR and TOT counters correctly.
  - Rearrange the order of the registers so that they are roughly
    in ascending order.
  - Reorder evcnt_attach_dynamic(), WM_EVCNT_ADD() and evcnt_detach()
    to match.
  - IC{TX,RX}*C registers are for older than 82575.
  - Fix a bug that the transmit underrun counter is incorrectly
    counted.
  - Don't add "Count" for event counter's description.
  - Some statistics registers were replaced with new counters on newer
    chips. Treat 0x403c(CEXTERR->HTDPMC), 0x40fc(TSCTFC->CBRMPC),
    0x4124(ICRXOC->HTCBDPC) and from 0x4104 to 0x4124.
  - Add some new counters:
    - Circuit Breaker TX Manageability Packet
    - Circuit Breaker RX Dropped Packet
    - Host Good Octets RX
    - Host Good Octets TX
    - Length Errors
    - SerDes/SGMII Code Violation Packet
    - Header Redirection Missed Packet
    - EEE TX LPI
    - EEE RX LPI
  - Fix prc511's comment and description.
- Add SOICZIFDATA (ifconfig -z) support for evcnt(9).
- Use WM_IS_ICHPCH(). No functional change.
- Fix typo. s/ictxact/ictxatc/. No functional change.
- Add comment.

Pull up following revision(s) (requested by msaitoh in ticket #216):

	sys/dev/pci/if_wm.c: revision 1.782

Add missing read to count Circuit Breaker Rx Dropped Packet correctly.

Add missing read to count Circuit Breaker Rx Dropped Packet correctly.

Pull up following revision(s) (requested by msaitoh in ticket #213):

	sys/dev/pci/if_wm.c: revision 1.770
	sys/dev/pci/if_wm.c: revision 1.771
	sys/dev/pci/if_wm.c: revision 1.772
	sys/dev/pci/if_wm.c: revision 1.773
	sys/dev/pci/if_wm.c: revision 1.774
	sys/dev/pci/if_wm.c: revision 1.775
	sys/dev/pci/if_wm.c: revision 1.776
	sys/dev/pci/if_wmreg.h: revision 1.129
	sys/dev/pci/if_wm.c: revision 1.777
	sys/dev/pci/if_wm.c: revision 1.778
	sys/dev/pci/if_wmvar.h: revision 1.49
	sys/dev/pci/if_wm.c: revision 1.779
	sys/dev/pci/if_wm.c: revision 1.768
	sys/dev/pci/if_wm.c: revision 1.769
	sys/dev/pci/if_wmreg.h: revision 1.130
	sys/dev/pci/if_wm.c: revision 1.780
	sys/dev/pci/if_wm.c: revision 1.781

Count some 64bit counters correctly.
 - Fix calculation of GORC, GOTC, TOR and TOT counters correctly.
 - Found by knakahara.

Add note for the TORL register.

The TOR register includes error, flow control and broadcast rejected.

Sort lines. No functional change.

 Rearrange the order of the registers so that they are roughly in ascending
order.

Sort lines. No functional change.

Reorder evcnt_attach_dynamic(), WM_EVCNT_ADD() and evcnt_detach() to match.
IC{TX,RX}*C registers are for older than 82575.

Fix a bug that the transmit underrun counter is incorrectly counted.
 The transmit underrun bit in the transmit status filed is only for 82544
(and older?), so don't use the counter for newer chips. The bit is reserved
for newer chips, but the bit sometimes set on 82575 at least.

Don't add "Count" for event counter's description.

 Some statistics registers were replaced with new counters.
 - 0x403c was CEXTERR(Carrier Extension Error). 82575 and newer except 80003,
   ICHs and PCHs have HTDPMC(Host Tx Discarded Packets by MAC). I don't really
   know for 82575. The 82575 datasheet say nothing about it.
 - The following two are changed for circuit breaker. Only 82576's datasheet
   describes abut it, so the registers might be only for 82576.
   Use those registers for 82575 and newer except 80003, ICHs and PCHs just in
   case.
   - 0x40fc was TSCTFC(TCP Segmentation Context Tx Fail). It was replaced by
     the CBRMPC(Circuit Breaker Rx Manageability Packet) register.
   - 0x4124 was ICRXOC(Interrupt Cause Receiver Overrun). It was replaced by
     the HTCBDPC(Host Tx Circuit Breaker Dropped Packet) register.
 - From 0x4104 to 0x4124:
   - For 0x4124, 82575's datasheet says it's not changed(ICRXOC).
     I don't know if it's correct. We use new HTCBDPC register for 82575.
   - 82576 and newer changed the meaning.
   - I don't know for 80003, ICHs and PCHs. Don't count those registers.
     At least, those registers in PCH2 and PCH_CNP are all zero.

Add some new event counters.

Add the following counters for 82575 and newer except 80003, ICHs and PCHs:
    - Only 82576 document describes about the circuit breaker,
      so the following two might be only for 82575:
        - Circuit Breaker TX Manageability Packet
        - Circuit Breaker RX Dropped Packet
    - 82575's document doesn't describe the following two, but we can see
      the same value as GO{T,R}C have:
        - Host Good Octets RX
        - Host Good Octets TX
    - 82575's document doesn't describe the LENERRS (Length Errors) counter.
      I don't know if it has.
    - Perhaps Non-SerDes/SGMII devices don't have SCVPC
      (SerDes/SGMII Code Violation Packet) register. We don't care if
      it's SerDes/SGMII or not for now.
    - HRMPC (Header Redirection Missed Packet) appears only once
      in 8257[56]'s datasheet. FreeBSD's igb counts it, so we do, too.
    - Count the following two for I350 and newer. I don't know if PCHs have:
        - EEE TX LPI
        - EEE RX LPI

 Move statistics updating code from wm_tick() to new wm_update_stats().
 - To reuse.
 - No functional change.

Add SOICZIFDATA (ifconfig -z) support for evcnt(9).

 First update the statistics data, then clear the event counters,
and finally copy and clear if_data via ether_ioctl().

Fix prc511's comment and description.

 Use WM_IS_ICHPCH(). No functional change.

Fix typo. s/ictxact/ictxatc/. No functional change.

Apply patch, requested by riastradh in ticket #1842:

	sys/dev/pci/if_wm.c			(apply patch)

Fix a bug introduced in ticket #1795.

Fix typo. s/ictxact/ictxatc/. No functional change.

 Use WM_IS_ICHPCH(). No functional change.

Fix prc511's comment and description.

Add SOICZIFDATA (ifconfig -z) support for evcnt(9).

 First update the statistics data, then clear the event counters,
and finally copy and clear if_data via ether_ioctl().

 Move statistics updating code from wm_tick() to new wm_update_stats().

 - To reuse.
 - No functional change.

Add some new event counters.

Add the following counters for 82575 and newer except 80003, ICHs and PCHs:
    - Only 82576 document describes about the circuit breaker,
      so the following two might be only for 82575:
        - Circuit Breaker TX Manageability Packet
        - Circuit Breaker RX Dropped Packet
    - 82575's document doesn't describe the following two, but we can see
      the same value as GO{T,R}C have:
        - Host Good Octets RX
        - Host Good Octets TX
    - 82575's document doesn't describe the LENERRS (Length Errors) counter.
      I don't know if it has.
    - Perhaps Non-SerDes/SGMII devices don't have SCVPC
      (SerDes/SGMII Code Violation Packet) register. We don't care if
      it's SerDes/SGMII or not for now.
    - HRMPC (Header Redirection Missed Packet) appears only once
      in 8257[56]'s datasheet. FreeBSD's igb counts it, so we do, too.
    - Count the following two for I350 and newer. I don't know if PCHs have:
        - EEE TX LPI
        - EEE RX LPI

 Some statistics registers were replaced with new counters.

 - 0x403c was CEXTERR(Carrier Extension Error). 82575 and newer except 80003,
   ICHs and PCHs have HTDPMC(Host Tx Discarded Packets by MAC). I don't really
   know for 82575. The 82575 datasheet say nothing about it.
 - The following two are changed for circuit breaker. Only 82576's datasheet
   describes abut it, so the registers might be only for 82576.
   Use those registers for 82575 and newer except 80003, ICHs and PCHs just in
   case.
   - 0x40fc was TSCTFC(TCP Segmentation Context Tx Fail). It was replaced by
     the CBRMPC(Circuit Breaker Rx Manageability Packet) register.
   - 0x4124 was ICRXOC(Interrupt Cause Receiver Overrun). It was replaced by
     the HTCBDPC(Host Tx Circuit Breaker Dropped Packet) register.
 - From 0x4104 to 0x4124:
   - For 0x4124, 82575's datasheet says it's not changed(ICRXOC).
     I don't know if it's correct. We use new HTCBDPC register for 82575.
   - 82576 and newer changed the meaning.
   - I don't know for 80003, ICHs and PCHs. Don't count those registers.
     At least, those registers in PCH2 and PCH_CNP are all zero.

Don't add "Count" for event counter's description.

Fix a bug that the transmit underrun counter is incorrectly counted.

 The transmit underrun bit in the transmit status filed is only for 82544
(and older?), so don't use the counter for newer chips. The bit is reserved
for newer chips, but the bit sometimes set on 82575 at least.

IC{TX,RX}*C registers are for older than 82575.

Sort lines. No functional change.

Reorder evcnt_attach_dynamic(), WM_EVCNT_ADD() and evcnt_detach() to match.

Sort lines. No functional change.

 Rearrange the order of the registers so that they are roughly in ascending
order.

Add note for the TORL register.

The TOR register includes error, flow control and broadcast rejected.

Count some 64bit counters correctly.

 - Fix calculation of GORC, GOTC, TOR and TOT counters correctly.
 - Found by knakahara.

Pull up the following revisions, requested by msaitoh in ticket #1795:

	sys/dev/pci/if_wm.c				1.764-1.767 via patch
	sys/dev/pci/if_wmreg.h				1.128

- Workaround for some hypervisor environments. The environments
  cannot stop e1000 interrupt immediately.
- Rename nq_txdesc' member "nqrx_ctx" to "nqtx_ctx". No functional
  change.
- Add comment. Modify comment.
- KNF.

Pull up the following revisions, requested by msaitoh in ticket #1578:

	sys/dev/pci/if_wm.c				1.764-1.767 via patch
	sys/dev/pci/if_wmreg.h				1.128

- Workaround for some hypervisor environments. The environments
  cannot stop e1000 interrupt immediately.
- Rename nq_txdesc' member "nqrx_ctx" to "nqtx_ctx". No functional
  change.
- Add comment. Modify comment.
- KNF.

Workaround for some hypervisor environments.

The environments cannot stop e1000 interrupt immediately.

Add comment to ICH/PCH's lock stuff in wm_reset().

Fix typo. No functional change. OK'd by knakahara.

Undo accidental whitespace changes from applying the patch for
ticket #1759 by applying a patch from msaitoh (NFC).

KNF. Modify comment. No fucntional change.

Pull up the following revisions, requested by msaitoh in ticket #1759:

	sys/dev/pci/if_wm.c				1.741-1.749,1.753-1.757,
							1.762 via patch
	sys/dev/pci/if_wmreg.h				1.126-1.127

- Fix I219 workaround in wm_flush_desc_rings().
- Add more statistics counters.
- To avoid releasing mutex temporally, use new
  wm_set_mdio_slow_mode_hv_locked().
- No functional changes:
  - Turn a locking botch (shouldn't drop lock on error) into a KASSERT
    in wm_deferred_start_locked().
  - Remove unneeded header inclusion.
  - Use __BIT() a little.
  - Modify comment and debug messages.
  - Consistency use -1 instead of 1 for some error code.
  - KNF.

Pull up the following revisions, requested by msaitoh in ticket #1515:

	sys/dev/pci/if_wm.c				1.741-1.749,
							1.753-1.757,
							1.762 via patch
	sys/dev/pci/if_wmreg.h				1.126-1.127

- Fix I219 workaround in wm_flush_desc_rings().
- Add more statistics counters.
- To avoid releasing mutex temporally, use new
  wm_set_mdio_slow_mode_hv_locked().
- No functional changes:
  - Turn a locking botch (shouldn't drop lock on error) into a KASSERT
    in wm_deferred_start_locked().
  - Remove unneeded header inclusion.
  - Use __BIT() a little.
  - Modify comment and debug messages.
  - Consistency use -1 instead of 1 for some error code.
  - KNF.

wm(4): Remove the non-MP-safe scaffolding.

- Where we had #ifndef WM_MPSAFE splnet(), we also had WM_CORE_LOCK,
  which implies splnet, so just remove the conditional splnet.

- Make the core lock unconditional and remove macro indirections.

- Pass CALLOUT_MPSAFE, SOFTINT_MPSAFE, and WQ_MPSAFE directly without
  macro indirections.

wm(4): Audit sc->phy.acquire and sc->nvm.acquire.

The return value must be used, because some of the acquire/release
pairs hold a mutex from acquire to release, and failing to call
release, or calling release twice, leads to an inconsistent state.

wm: use a workqueue to reset the interface when wm_watchdog determines it needs a reset.


Author: Nick Hudson <skrll@netbsd.org>
Committer: Taylor R Campbell <riastradh@NetBSD.org>

wm(4): if_flags and IFNET_LOCK audit

Don't touch if_flags without IFNET_LOCK:

- If only core lock is held, use sc_if_flags.
- If only txq lock is held, use txq_stopping.
  => Verified all paths guarantee !txq_stopping, so assert.
- Make sure sc_if_flags is updated on stop.
- Make wm_init fail once we enter wm_detach.
- Sprinkle assertions.


Author: Taylor R Campbell <riastradh@NetBSD.org>

wm(4): Revert previous -- mistakenly committed unsquashed change part.

wm(4): if_flags and IFNET_LOCK audit

Don't touch if_flags without IFNET_LOCK:

- If only core lock is held, use sc_if_flags.
- If only txq lock is held, use txq_stopping.
  => Verified all paths guarantee !txq_stopping, so assert.
- Make sure sc_if_flags is updated on stop.
- Sprinkle assertions.

To avoid releasing mutex temporally, use new wm_set_mdio_slow_mode_hv_locked().

Consistency use -1 instead of 1 for some error code. Advised by knakahara.

Pass an error code correctly if phy.acquire() failed.

This is not a real bug because the return value is not used.

Pass an error code to the upper layer instead of -1. Advised by knakahara.

Modify debug messages so that we can determine where it happened.

Sprinkle const on splfoo call results.

Add a KASSERT for the locking protocol in wm_ioctl.

Read the interface up/down status from sc_if_flags (under WM_CORE_LOCK)
when deciding if the multicast filter needs to be updated.

Discussed with msaitoh@, knakahara@ and riastradh@

Add some KASSERTs around the locking protocol.

Discussed with msaitoh@, knakahara@ and riastradh@

Improve a couple of comments

Trailing whitespace

KNF the KNF

Add more statistics countes.

 - Add many statics counters that the chip has.
 - Attach event counters only if available.

wm(4): Use CSR_WRITE_FLUSH, not bus_space_barrier.

The bus space is mapped non-prefetchable and non-cacheable, so there
is no need for bus_space_barrier and it might not even have any
effect -- bus_space_read/write are totally ordered for non-
prefetchable, non-cacheable mappings anyway.  Presumably the intent
here was to wait for acknowledgment of the write from the device --
which bus_space_barrier does not guarantee, and which I expect
requires a read transaction as CSR_WRITE_FLUSH performs.

ok msaitoh@, knakahara@

wm(4): Use bus_dmamap_sync, not bus_space_barrier, for tx desc.

The newly written tx descriptor needs to be visible to the device
before the TDT register is set.  There's no bus space mappings that
need any barriers, though -- nothing mapped prefetchable or cacheable
here.  So bus_dmamap_sync (via wm_cdtxsync) is necessary, not
bus_space_barrier.  By accident, bus_space_barrier implies the same
barrier instructions as bus_dmamap_sync on some architectures, but it
wouldn't work at all if we were bouncing.

ok msaitoh@, knakahara@

KNF

Don't need sys/malloc.h so remove the #include

Turn a locking botch (shouldn't drop lock on error) into a KASSERT

Pull up the following revisions, requested by msaitoh in ticket #1751:

	sys/dev/pci/if_wmreg.h				1.122-1.125 via patch
	sys/dev/pci/if_wmvar.h				1.48
	sys/dev/pci/if_wm.c				1.719-1.720,1.722-1.725,
							1.727-1.740 via patch

- wm_tick: Add missing splx(s) when not WM_MPSAFE.
- Print DMA range info if the system is booting in the verbose mode.
- Micro optimization:
  - Call m_freem(m) only if m0 == NULL.
  - Call wm_xxeof() only when limit > 0.
  - Don't set the more flag when there is no packet to process.
- No functional changes:
  - Call txeof first, then rxeof for the consistency.
  - Remove duplicated break.
  - Remove stray semicolons from struct declaration.
  - Fix value return from void function.
  - Use macros.
  - Modify comment.
  - KNF.

Pull up the following revisions, requested by msaitoh in ticket #1477:

	sys/dev/pci/if_wmreg.h				1.122-1.125
	sys/dev/pci/if_wmvar.h				1.48
	sys/dev/pci/if_wm.c				1.719-1.720,
							1.722-1.725,
							1.727-1.740 via patch

- wm_tick: Add missing splx(s) when not WM_MPSAFE.
- Print DMA range info if the system is booting in the verbose mode.
- Micro optimization:
  - Call m_freem(m) only if m0 == NULL.
  - Call wm_xxeof() only when limit > 0.
  - Don't set the more flag when there is no packet to process.
- No functional changes:
  - Call txeof first, then rxeof for the consistency.
  - Remove duplicated break.
  - Remove stray semicolons from struct declaration.
  - Fix value return from void function.
  - Use macros.
  - Modify comment.
  - KNF.

KNF. No functional change.

Consistently use semicolon. No functional change.

Call txeof first, then rxeof for the consistency.

 There are three functions where the txeof and rxeof are called. The MSI-X
interrupt function and the workqueue function call rxeof first, then rxeof.
For legacy interrupt. rxeof is called first. Modify it to match with other
two.

Fix compile error if WM_DEBUG is set.

Be less tautological and more explanatory in a comment.

Use PRIxBIT instead of casts

wm_tick: Add missing splx(s) when not WM_MPSAFE

Don't set the more flag when there is no packet to process.

 Move the location of the loop limit check next to the DD bit test. This
change prevent from scheduling the next txrx_work that does nothing.

Call wm_xxeof() only when limit > 0.

 It's not required to call wm_xxeof() when limit == 0. It means that
all xxeof processing will be done by softint or workqueue. Currently,
wm_xxeof() returns quickly before checking the descriptor head when
limit == 0 and return with ture to set the more flag.

Call m_freem(m) only if m0 == NULL.

if_wm.c: various spelling and grammar fixes

if_wm.c: fix value return from void function

lint complained:
if_wm.c(10028): error:
    void function wm_txrxintr_disable cannot return value [213]
if_wm.c(10049): error:
    void function wm_txrxintr_enable cannot return value [213]

No functional change.

if_wm.c: remove stray semicolons from struct declaration

C99 does not allow 'struct { int member;; }'.

No binary change.

Remove duplicated break. No functional change.

sys: Use if_init wrapper function.

Exception: Not in kern_pmf.c, for the kind of silly reason that it
avoids having kern_pmf.c refer to symbols defined only in net; this
avoids a pain in the rump.

Keep constants 32 bit, why does __BIT() return uintmax_t?

aprint_verbose the DMA range used.

whitespace in a comment

Typo in comment

Simplify.  Same code before and after.

Trailing whitespace

s/checkum/checksum/ in comment.

Pull up the following, requested by msaitoh in ticket #1707:

	sys/dev/pci/if_wm.c				1.716-1.718 via patch
	sys/dev/pci/if_wmreg.h				1.121
	sys/dev/mii/ihphy.c				1.20

- Fix a bug that device timeout still happens when the link is down
  on ICH/PCH. Fixes PR kern/56478.
- Add some sysctl info for debugging.
- ihphy(4): Don't power down the PHY when the interface goes down.
- Use macro. Fix comment.

Pull up the following, requested by msaitoh in ticket #1373:

	sys/dev/pci/if_wm.c				1.716-1.718 via patch
	sys/dev/pci/if_wmreg.h				1.121
	sys/dev/mii/ihphy.c				1.20

- Fix a bug that device timeout still happens when the link is down
  on ICH/PCH. Fixes PR kern/56478.
- Add some sysctl info for debugging.
- ihphy(4): Don't power down the PHY when the interface goes down.
- Use macro. Fix comment.

Fix a bug that device timeout still happens on ICH/PCH. Fixes PR kern/56478.

 - if_wm.c rev. 1.695 for PR kern/40981 still had a problem. The workaround
   flag was unexpectedly cleared when a cable is not connected, SIOCINITIFADDR
   was called, or if_init was called. Fix it.
 - Add debug printf()s.

 Use macro. No functional change.

Add some sysctl info for debugging.

Pull up the following, requested by msaitoh in ticket #1701:

	share/man/man4/wm.4				1.42
	sys/dev/pci/files.pci				1.434
	sys/dev/pci/if_wm.c				1.686-1.687,1.692-1.693,
							1.697,1.699-1.703,
							1.706-1.715 via patch
	sys/dev/pci/if_wmvar.h				1.46

- Add missing drain for pcq in wm_stop_locked().
- Add support for I219V 15-19 and I219LM 16-19.
- Fix Tx stall.
- Use wm_flush_desc_rings() workaround more on I219.
- Change DMA physical address in wm_flush_desc_rings() to match other
  OSes.
- Check return value correctly in wm_lv_jumbo_workaround_ich8lan().
- Add new sysctl hw.wmN.debug_flags. This sysctl can be used if
  WM_DEBUG is set.
- Add some sysctl values for debugging TX/RX queues.
- WM_EVENT_COUNTER is enabled by default on 64 bit architectures.
- Remove extra unlock/lock processing around if_percpuq_enqueue().
- Refactor rxq->rxq_ptr updating.
- Stop legacy interrupts before calling softint.
- Disable printf()s in wm_flush_desc_rings() because the code is
  verified.
- Print I219's version number.
- Uniform INTx/MSI handler's Tx/Rx behavior to MSI-X's one.
- Fix return value of interrupt handler.
- Only print an error about missing I/O BARs for chips that need it.
- Do not return a void value from a void function.
- Add new flag named WM_F_CRC_STRIP and use it. No functional change.
- Whitespace fixes. Fix comments. No functional change.

Pull up the following, requested by msaitoh in ticket #1363:

	share/man/man4/wm.4			1.42
	sys/dev/pci/files.pci			1.434
	sys/dev/pci/if_wm.c			1.690,1.692-1.693,1.697-1.704,
						1.706-1.715 via patch

- Add missing drain for pcq in wm_stop_locked().
- Add support for I219V 15-19 and I219LM 16-19.
- Fix Tx stall.
- Use wm_flush_desc_rings() workaround more on I219.
- Change DMA physical address in wm_flush_desc_rings() to match other
  OSes.
- Check return value correctly in wm_lv_jumbo_workaround_ich8lan().
- Add new sysctl hw.wmN.debug_flags. This sysctl can be used if
  WM_DEBUG is set.
- Add some sysctl values for debugging TX/RX queues.
- Use atomic_{load,store}_relaxed for evcnt 64 bit counter.
- WM_EVENT_COUNTER is enabled by default on 64 bit architectures.
- Remove extra unlock/lock processing around if_percpuq_enqueue().
- Refactor rxq->rxq_ptr updating.
- Stop legacy interrupts before calling softint.
- Disable printf()s in wm_flush_desc_rings() because the code is
  verified.
- Print I219's version number.
- Uniform INTx/MSI handler's Tx/Rx behavior to MSI-X's one.
- Fix return value of interrupt handler.
- Only print an error about missing I/O BARs for chips that need it.
- Do not return a void value from a void function.
- Fix comment.
- Whitespace fixes. No functional change.

Use wm_flush_desc_rings() more.

 - Use wm_flush_desc_rings() for newer than PCH_SPT, too. Same as other OSes.
   It seems that some devices has no this errata, but we don't know how to
   identify it. So just use >= PCH_SPT.
 - Use wm_flush_desc_rings() before chip reset in wm_resume(), too.

Disable printf()s in wm_flush_desc_rings() because the code is verified.

 The problem was very rare, so I added those printf()s as we can see.
Last week, knakahara and I found a procedure to reproduce the problem and
verified it worked correctly.

 To reproduce the problem:

 0) Use I219 V1 to V5. Not all of them have the problem and newer than V5 may
    have the problem. I used V2 for the test.
 1) Define WM_DEBUG or add printf in wm_flush_desc_rings() to see
    DESCRING_STATUS_FLUSH_REQ bit.
 2) Run "iperf -s" on the DUT.
 3) Run "iperf3 -R -u -c 192.168.1.110 -w 8m -b 1g" repeatedly on another
    machine. Note that kern.sbmax should be increased on both machines.
 4) run
 	while true; do
	ifconfig wm0 down up
	sleep 30
	done
 5) After DESCRING_STATUS_FLUSH_REQ is set, without wm_flush_desc_rings(),
    TX stalls and reboot is required to recover from it.
    With wm_flush_desc_rings(), no device hang.
 6) If you can't see the DESCRING_STATUS_FLUSH_REQ is set, change the media
    to 100BASE-T.

 Note that if_wm.c rev. 1.710 originally written by chuq is required for the
test. Without the change, the device becomes unrecoverable state before
wm_flush_desc_rings() and the function doesn't recover from the problem.

Change DMA physical address in wm_flush_desc_rings() to match other OSes.

 - Even without this change, the workaround works with my machines.
 - Add comment.

Fix return value of interrupt handler, pointed out by msaitoh@n.o.

Even if stopping flag is set (and handler do nothing), the interrupt
itself should be assume to be processed when ICR is not zero.  If not,
interrupt count is not incremented correctly.

Uniform INTx/MSI handler's Tx/Rx behavior to MSI-X's one.

Because the difference has caused INTx/MSI own bugs.

Fix Tx stall, reported and implemented original patch by chs@n.o, thanks.

Check if_snd queue because if.c::if_transmit() doesn't call ifp->if_start()
when IFQ_ENQUEUE failed.

Reviewed and tested by chs@n.o and msaitoh@n.o.

Only print an error about missing I/O BARs for chips that need it.

Add support for I219V 15-19 and I219LM 16-19.

Print I219's version number.

Fix locking against myself reported by rin@n.o.

In some machine, INTx interrupts occurs for some reason while holding
txq_lock in softint context.
e.g. https://dmesgd.nycbug.org/index.cgi?do=view&id=6139

The panic message is the following.
====================
[  50.1699480] panic: LOCKDEBUG: Mutex error: mutex_vector_enter,467: locking against myself
[  50.1800188] cpu0: Begin traceback...
[  50.1800188] 0xc7a29ac4: netbsd:db_panic+0xc
[  50.1900018] 0xc7a29adc: netbsd:vpanic+0x120
[  50.1999533] 0xc7a29af4: netbsd:aprint_debug
[  50.1999533] 0xc7a29b34: netbsd:lockdebug_abort1+0x18c
[  50.2099719] 0xc7a29b74: netbsd:lockdebug_wantlock+0x1e8
[  50.2199525] 0xc7a29bb4: netbsd:mutex_enter+0x288
[  50.2299642] 0xc7a29be4: netbsd:wm_intr_legacy+0x9c
[  50.2299642] 0xc7a29c2c: netbsd:i80321_intr_dispatch+0x138
[  50.2399631] 0xc7a29de4: netbsd:irq_entry+0x98
[  50.2499452] 0xc7a29e24: netbsd:entropy_enter+0x88
[  50.2599996] 0xc7a29e6c: netbsd:rnd_add_data_1+0x50
[  50.2699565] 0xc7a29ebc: netbsd:rnd_add_data+0x10c
[  50.2699565] 0xc7a29ed4: netbsd:rnd_add_uint32+0x28
[  50.2799524] 0xc7a29f1c: netbsd:wm_txeof+0x2c0
[  50.2899527] 0xc7a29f5c: netbsd:wm_handle_queue+0x44
[  50.3000257] 0xc7a29fac: netbsd:softint_thread+0x130
[  50.3100119] cpu0: End traceback...
====================

To fix this panic, stop wm(4) interrupts before calling softint.
This behavior is similar to MSI-X polling mode.

Reviewed by msaitoh@n.o, tested by rin@n.o, thanks.

Sync w/ HEAD.

if_attach and if_initialize cannot fail, don't test return value

These were originally made failable back in 2017 when if_initialize
allocated a softint in every interface for link state changes, so
that it could fail gracefully instead of panicking:

https://mail-index.NetBSD.org/source-changes/2017/10/23/msg089053.html

However, this spawned many seldom- or never-tested error branches,
which are risky to have around.  And that softint in every interface
has since been replaced by a single global workqueue, because link
state changes require thread context but not low latency or high
throughput:

https://mail-index.NetBSD.org/source-changes/2020/02/06/msg113759.html

So there is no longer any reason for if_initialize to fail.  (The
subroutine if_stats_init can't fail because percpu_alloc can't fail
either.)

There is a snag: the softint_establish in if_percpuq_create could
fail, potentially leading to bad consequences later on trying to use
the softint.  This change doesn't introduce any new bugs because of
the snag -- if_percpuq_attach was already broken.  However, the snag
can be better addressed without spawning error branches, either by
using a single softint or making softints less scarce.

(Separate commit will change the signatures of if_attach and
if_initialize to return void, scheduled to ride whatever is the next
convenient kernel bump.)

Patch and testing on amd64 and evbmips64-eb by maya@; commit message
soliloquy, and compile-testing on evbppc/i386/earmv7hf, by me.

sync with head

Sync with HEAD.

Fix i386 build failure with options WM_EVENT_COUNTERS.

pointed out by msaitoh@n.o.

if_wm: do not return a void value from a void function

This is a GCC extension.

Sync with HEAD.

Whitespace fixes. No functional change.

Refactor rxq->rxq_ptr updating.

Remove extra unlock/lock processing around if_percpuq_enqueue().

This temporary unlock/lock processing was required for direct calling
ifp->if_input.  After r1.391, wm(4) uses if_percpuq_enqueue() and it
does not hold any mutexes.

In 64 bit architectures, WM_EVENT_COUNTER is enabled by default.

No objection from tech-kern@n.o and tech-net@n.o.

ok'ed by msaitoh@n.o.

wm(4) use atomic_{load,store}_relaxed for evcnt 64 bit counter.

ok'ed by msaitoh@n.o.

Sync w/ HEAD.

 Add some sysctl values for debugging TX/RX queues.

Pull up following revision(s) (requested by rin in ticket #1133):

	sys/dev/pci/if_wm.c: revision 1.696

Fix little-endian dependence in wm_rxeof(), by which packets cannot be
received by 82574 and successors on big-endian machines.

Tested by aarch64eb with I210-T1 on ROCKPro64.

Thanks msaitoh for discussion!
XXX
pullup to netbsd-9 and netbsd-8

Pull up following revision(s) (requested by rin in ticket #1623):

	sys/dev/pci/if_wm.c: revision 1.696

Fix little-endian dependence in wm_rxeof(), by which packets cannot be
received by 82574 and successors on big-endian machines.

Tested by aarch64eb with I210-T1 on ROCKPro64.

Thanks msaitoh for discussion!
XXX
pullup to netbsd-9 and netbsd-8

Fix little-endian dependence in wm_rxeof(), by which packets cannot be
received by 82574 and successors on big-endian machines.

Tested by aarch64eb with I210-T1 on ROCKPro64.

Thanks msaitoh for discussion!

XXX
pullup to netbsd-9 and netbsd-8

Pull up following revision(s) (requested by knakahara in ticket #1620):

	sys/dev/pci/if_wm.c: revision 1.694
	sys/dev/pci/if_wm.c: revision 1.695 (via patch)
	sys/dev/pci/if_wmvar.h: revision 1.47

Add WMPHY_I350. Not used yet.

Workaround for ihphy and atphy(ICH*/PCH*, 82580 and I350).
These phys stop DMA while link is down which causes device timeout.
Fix PR/kern 40981
Reviewed and tested by msaitoh@n.o, thanks.
XXX pullup-[89]

Pull up following revision(s) (requested by knakahara in ticket #1126):

	sys/dev/pci/if_wm.c: revision 1.694
	sys/dev/pci/if_wm.c: revision 1.695 (via patch)
	sys/dev/pci/if_wmvar.h: revision 1.47

Add WMPHY_I350. Not used yet.

Workaround for ihphy and atphy(ICH*/PCH*, 82580 and I350).
These phys stop DMA while link is down which causes device timeout.
Fix PR/kern 40981
Reviewed and tested by msaitoh@n.o, thanks.
XXX pullup-[89]

Workaround for ihphy and atphy(ICH*/PCH*, 82580 and I350).

These phys stop DMA while link is down which causes device timeout.
Fix PR/kern 40981

Reviewed and tested by msaitoh@n.o, thanks.

XXX pullup-[89]

Add WMPHY_I350. Not used yet.

 Add new sysctl hw.wmN.debug_flags. This sysctl can be used if WM_DEBUG is set.
OK'd by knakahara.

% sysctl -d hw.wm0.debug_flags
hw.wm0.debug_flags: Debug flags:
        0x01 LINK
        0x02 TX
        0x04 RX
        0x08 GMII
        0x10 MANAGE
        0x20 NVM
        0x40 INIT
        0x80 LOCK

 Add missing drain for pcq in wm_stop_locked(). OK'd by knakahara.

This change fixes two problems:
 1. If the pcq is full and watchdog timer is fired, the full state of the
    pcq is kept and wm_transmit() returns with ENOBUFS.
 2. ifconfig down doesn't free mbufs in the pcq.

Pull up following revision(s) (requested by msaitoh in ticket #1617):

	sys/dev/pci/if_wm.c: revision 1.691

  Fixes a problem that the attach function reported
"wm_gmii_setup_phytype: Unknown PHY model. OUI=000000, model=0000" and
"PHY type is still unknown." Don't call wm_gmii_setup_phytype() three times if
the interface uses SGMII with internal MDIO.

  Tested with I354(Rangeley(SGMII(MDIO))) and I350(SERDES(SFP), SGMII(SFP)).

Pull up following revision(s) (requested by msaitoh in ticket #1114):

	sys/dev/pci/if_wm.c: revision 1.691

  Fixes a problem that the attach function reported
"wm_gmii_setup_phytype: Unknown PHY model. OUI=000000, model=0000" and
"PHY type is still unknown." Don't call wm_gmii_setup_phytype() three times if
the interface uses SGMII with internal MDIO.

  Tested with I354(Rangeley(SGMII(MDIO))) and I350(SERDES(SFP), SGMII(SFP)).

 Fixes a problem that the attach function reported
"wm_gmii_setup_phytype: Unknown PHY model. OUI=000000, model=0000" and
"PHY type is still unknown." Don't call wm_gmii_setup_phytype() three times if
the interface uses SGMII with internal MDIO.

 Tested with I354(Rangeley(SGMII(MDIO))) and I350(SERDES(SFP), SGMII(SFP)).

Small fixes from knakahara:

- Check return value correctly in wm_lv_jumbo_workaround_ich8lan().
- Fix comment.

Pull up following revision(s) (requested by msaitoh in ticket #1093):

	sys/dev/pci/if_wmreg.h: revision 1.120
	sys/dev/pci/if_wmvar.h: revision 1.46
	sys/dev/pci/if_wm.c: revision 1.686
	sys/dev/pci/if_wm.c: revision 1.687
	sys/dev/mii/inbmphyreg.h: revision 1.20
	sys/dev/pci/if_wm.c: revision 1.688
	sys/dev/pci/if_wm.c: revision 1.689

s/ressource/resource/. Found by knakahara.

  Add new flag named WM_F_CRC_STRIP and use it. No functional change.

  This change also sets the RCTL_SECRC bit on I211 but it doesn't change
the behavior because I211 always strips CRC like I35[04] and I210.

  Add a workaround for jumbo frame on PCH2 and newer. Tested by chs@.

- Add wm_lv_jumbo_workaround_ich8lan() and use it. From FreeBSD.
   XXX For KUMCTRLSTA_OFFSET_HD_CTRL register modification, it's doubtful.
   FreeBSD and Linux do the same thing that they set the same value on both
   jumbo frame's enable case and the disable case. It seems the default value
   is 0x0b0c and it's not changed on the enable case, so it might be a bug
   on the enable case or the modification is not required.
- Rename I219_UNKNOWN1 to I82579_UNKNOWN1.
  Use 12K for packet buffer for jumbo frame on PCH2 and newer.

XXX Note that Linux Use 14K.

 Use 12K for packet buffer for jumbo frame on PCH2 and newer.

XXX Note that Linux Use 14K.

 Add a workaround for jumbo frame on PCH2 and newer. Tested by chs@.

- Add wm_lv_jumbo_workaround_ich8lan() and use it. From FreeBSD.

  XXX For KUMCTRLSTA_OFFSET_HD_CTRL register modification, it's doubtful.
  FreeBSD and Linux do the same thing that they set the same value on both
  jumbo frame's enable case and the disable case. It seems the default value
  is 0x0b0c and it's not changed on the enable case, so it might be a bug
  on the enable case or the modification is not required.

- Rename I219_UNKNOWN1 to I82579_UNKNOWN1.

 Add new flag named WM_F_CRC_STRIP and use it. No functional change.

 This change also sets the RCTL_SECRC bit on I211 but it doesn't change
the behavior because I211 always strips CRC like I35[04] and I210.

s/ressource/resource/. Found by knakahara.

Pull up the following revisions, requested by msaitoh in ticket #1594:

	sys/dev/pci/if_wm.c			1.655-1.658, 1.660,
						1.662, 1.664-1.668,
						1.671-1.674, 1.678,
						1.680-1.685 via patch
	sys/dev/pci/if_wmreg.c			1.118-1.119 via patch
	sys/dev/pci/if_wmvar.c			1.45 via patch
	sys/dev/mii/igphy.c			1.35-1.36 via patch
	sys/dev/mii/igphyreg.h			1.12-1.13
	sys/dev/mii/makphy.c			1.66 via patch
	sys/dev/mii/makphyreg.h			1.11

- Add SFP support. Module insertion/removal is not supported yet.
  Currently, SFP detection is only done in the driver's attach phase.
- Detect the Media Auto Sense feature. Not supported yet.
- Fix SFF_SFP_ETH_FLAGS_100FX. It's not 0x10 but 0x20.
- Add extra delay in wm_serdes_power_up_link_82575().
- Add Intel I219 LM10-LM15 and V10-V14.
- wm(4) can use workqueue as deferred Rx/Tx handler).
  Set hw.wm*.txrx_workqueue=1 to use workqueue instead of softint.
  The default value of hw.wm*.txrx_workqueue is 0 which use softint
  as before.
- Unset RSS UDP flags like ixg(4) and other OSes. To handle IP
  fragmented UDP, first packet and second packet should be processed
  in the same Rx queue.
- It's useless to not to set PCI_PMCSR_PME_STS bit when writing because
  the bit is W1C. Instead, always write PCI_PMCSR_PME_STS bit to clear
  in case it's already set.
- Actually writing always the checksum offload context descriptor
  makes the HW do extra processing, avoid doing that if possible.
- Fix a bug that the WMREG_EEARBC_I210 register is incorrectly set if
  the system uses iNVM.
- "wmX: 0" on 82542 is difficult to understand, so don't print it.
- Explicitly cast from uint16_t to uint32_t before shifting 16bit left
  when printing Image Unique ID to avoid undefined behavior.
- Set if_baudrate for non-MII device.
- Rename some macros and function.
- KNF. Add comment.

Pull up following revision(s) (requested by msaitoh in ticket #1040):

	sys/dev/mii/igphy.c: revision 1.35
	sys/dev/mii/igphy.c: revision 1.36
	sys/dev/mii/igphyreg.h: revision 1.12
	sys/dev/mii/igphyreg.h: revision 1.13
	sys/dev/mii/makphyreg.h: revision 1.11
	sys/dev/pci/if_wm.c: revision 1.682
	sys/dev/pci/if_wm.c: revision 1.683
	sys/dev/pci/if_wm.c: revision 1.684
	sys/dev/pci/if_wm.c: revision 1.685
	sys/dev/mii/makphy.c: revision 1.66

s/MII_IGPHY_/IGPHY_/. No functional change.

Rename PSSR_* to MAKPHY_PSSR_* and IGPHY_PSSR_* to avoid conflict.
No functional change.

  Setup PCS and SGMII for SFP correctly. It still doesn't support SFP
insertion/removal.

Copper:
wm2: SGMII(SFP)
wm2: 0x1043c440<SPI,IOH_VALID,PCIE,SGMII,NEWQUEUE,ASF_FIRM,EEE,SFP>
makphy0 at wm2 phy 6: Marvell 88E1111 Gigabit PHY, rev. 1

Fiber:
wm3: SERDES(SFP)
wm3: 0x10034440<SPI,IOH_VALID,PCIE,NEWQUEUE,ASF_FIRM,SFP>
wm3: 1000baseSX, 1000baseSX-FDX, auto

  Explicitly cast from uint16_t to uint32_t before shifting 16bit left
when printing Image Unique ID. Found by kUBSan.

  Set if_baudrate for non-MII device. Before this commit, it was 0.

 Set if_baudrate for non-MII device. Before this commit, it was 0.

 Explicitly cast from uint16_t to uint32_t before shifting 16bit left
when printing Image Unique ID. Found by kUBSan.

 Setup PCS and SGMII for SFP correctly. It still doesn't support SFP
insertion/removal.

Copper:
wm2: SGMII(SFP)
wm2: 0x1043c440<SPI,IOH_VALID,PCIE,SGMII,NEWQUEUE,ASF_FIRM,EEE,SFP>
makphy0 at wm2 phy 6: Marvell 88E1111 Gigabit PHY, rev. 1

Fiber:
wm3: SERDES(SFP)
wm3: 0x10034440<SPI,IOH_VALID,PCIE,NEWQUEUE,ASF_FIRM,SFP>
wm3: 1000baseSX, 1000baseSX-FDX, auto

s/MII_IGPHY_/IGPHY_/. No functional change.

Pull up the following revisions, requested by msaitoh in ticket #994:

	sys/dev/pci/if_wm.c		1.655-1.658, 1.660, 1.662, 1.664-1.668,
					1.671-1.674, 1.678,1.680-1.681 via patch
	sys/dev/pci/if_wmreg.c		1.118-1.119
	sys/dev/pci/if_wmvar.c		1.45

- Add SFP support. Module insertion/removal is not supported yet.
  Currently, SFP detection is only done in the driver's attach phase.
- Detect the Media Auto Sense feature. Not supported yet.
- Fix SFF_SFP_ETH_FLAGS_100FX. It's not 0x10 but 0x20.
- Add extra delay in wm_serdes_power_up_link_82575().
- Add Intel I219 LM10-LM15 and V10-V14.
- wm(4) can use workqueue as deferred Rx/Tx handler.
  Set hw.wm*.txrx_workqueue=1 to use workqueue instead of softint.
  The default value of hw.wm*.txrx_workqueue is 0 which use softint
  as before.
- Unset RSS UDP flags like ixg(4) and other OSes. To handle IP
  fragmented UDP, first packet and second packet should be processed
  in the same Rx queue.
- It's useless to not to set PCI_PMCSR_PME_STS bit when writing because
  the bit is W1C. Instead, always write PCI_PMCSR_PME_STS bit to clear
  in case it's already set.
- Actually writing always the checksum offload context descriptor
  makes the HW do extra processing, avoid doing that if possible.
- Fix a bug that the WMREG_EEARBC_I210 register is incorrectly set if
  the system uses iNVM.
- "wmX: 0" on 82542 is difficult to understand, so don't print it.
- Rename some macros and function.
- KNF. Add comment.

"wmX: 0" on 82542 is difficult to understand, so don't print it.

 Fix a bug that the WMREG_EEARBC_I210 register is incorrectly set if the system
uses iNVM.

prop_data_data_nocopy -> prop_data_value

No functional change:

 - Rename some macros and function.
 - Add newline.

Update for proplib(3) API changes.

Make WM_NRXDESC unsigned.

Revert "- rnd_add_*() call must be serialized. Unserialized accesses broke the"

It is no longer necessary to serialize calls to rnd_add_data, so we
can use the same rndsource for every queue.

This also reverts some unrelated changes that were mixed in:

1. WM_NRXDESC was made unsigned
2. The specific samples were changed.

(1) will be redone in a separate commit.  (2) was buggy (wrote `='
where `|=' was meant) and is not needed; it's fine to do rnd_add_data
in a thread or softint too, not just in a hard interrupt handler.

Sync with HEAD

Mostly merge changes from HEAD upto 20200411

adjust the condition for reusing checksum context, wm_tx_offload() can
be used even for multiqueue case

enable only when the chip supports exactly one queue (same as FreeBSD) -
anything multiqueue-capable later than 82574 should work, but I don't
have the hw to test

wm_tx_offload() and wm_nq_tx_offload() actually always return 0, make them
return void and remove the check for their return value

fixup field names in previous

actually writing always the checksum offload context descriptor
makes the hw do extra processing, avoid doing that if possible -
on my computer with I219 this results in about 2% speedup in Tx performance

change adoped from FreeBSD

XXX should be also done for multiqueue case, but I don't have hw to test it

Don't bother with IFF_OACTIVE.

Define and implement a locking protocol for the ifmedia / mii layers:
- MP-safe drivers provide a mutex to ifmedia that is used to serialize
  access to media-related structures / hardware regsiters.  Converted
  drivers use the new ifmedia_init_with_lock() function for this.  The
  new name is provided to ease the transition.
- Un-converted drivers continue to call ifmedia_init(), which will supply
  a compatibility lock to be used instead.  Several media-related entry
  points must be aware of this compatibility lock, and are able to acquire
  it recursively a limited number of times, if needed.  This is a SPIN
  mutex with priority IPL_NET.
- This same lock is used to serialize access to PHY registers and other
  MII-related data structures.

The PHY drivers are modified to acquire and release the lock, as needed,
and assert the lock is held as a diagnostic aid.

The "usbnet" framework has had an overhaul of its internal locking
protocols to fit in with the media / mii changes, and the drivers adapted.

USB wifi drivers have been changed to provide their own adaptive mutex
to the ifmedia later via a new ieee80211_media_init_with_lock() function.
This is required because the USB drivers need an adaptive mutex.

Besised "usbnet", a few other drivers are converted: vmx, wm, ixgbe / ixv.

mcx also now calls ifmedia_init_with_lock() because it needs to also use
an adaptive mutex.  The mcx driver still needs to be fully converted to
NET_MPSAFE.

Sync with head.

 It's useless to not to set PCI_PMCSR_PME_STS bit when writing because
the bit is W1C. Instead, always write PCI_PMCSR_PME_STS bit to clear in
case it's already set.

 Whitespace fix. No functional change.

wm(4) unset RSS UDP flags like ixg(4) and other OS's one.

To handle IP fragmented UDP first packet and second packet by the same
Rx queue.

I tested on I354.

Fix missing workqueue_destroy().  ok by msaitoh@n.o

Fix wm(4) create too many workqueue.  Pointed out by msaitoh@n.o

Adopt <net/if_stats.h>.

Pull up the following (via patch), requested by msaitoh in ticket #650:

	sys/dev/pci/if_wm.c			1.650, 1.652-1.654
	sys/dev/pci/if_wmreg.h			1.116-1.117

- Set CTRL_ILOS(Invert loss of signal) bit correctly on 82580
  port 1, 2, 3 and newer chips. This change fixes a bug that some
  fiber, serdes or SFP devices don't detect the link status correctly.
- Simplify code by using "struct mii_data *mii" more. No functional
  change.
- MSI-X doesn't use sc->sc_icr variable, so move the code into
  non-MSI-X part. No functional change intended.
- Modify debug printfs a bit.
- Rename macro.
- Use __BIT()
- Fix comment. Add comment.
- KNF.

Sync with head.

Pull up the following, requested by msaitoh in ticket #1492:

	sys/dev/pci/if_wm.c			1.650, 1.652-1.654 via patch
	sys/dev/pci/if_wmreg.h			1.116-1.117

- Set CTRL_ILOS(Invert loss of signal) bit correctly on 82580
  port 1, 2, 3 and newer chips. This change fixes a bug that some
  fiber, serdes or SFP devices don't detect the link status correctly.
- Simplify code by using "struct mii_data *mii" more. No functional
  change.
- MSI-X doesn't use sc->sc_icr variable, so move the code into
  non-MSI-X part. No functional change intended.
- Modify debug printfs a bit.
- Rename macro.
- Use __BIT()
- Fix comment. Add comment.
- KNF.

wm(4) can use workqueue as deferred Rx/Tx handler like ixg(4).

Set hw.wm*.txrx_workqueue=1 to use workqueue instead of softint.
The default value of hw.wm*.txrx_workqueue is 0, that is, use softint
as before.

ok by msaitoh@n.o.

Pull up following revision(s) (requested by knakahara in ticket #1488):

	sys/dev/pci/if_wm.c: revision 1.661

fix softint_disestablish() in wm_detach().

ok by msaitoh@n.o
XXX pullup-8, pullup-9

Pull up following revision(s) (requested by knakahara in ticket #643):

	sys/dev/pci/if_wm.c: revision 1.661

fix softint_disestablish() in wm_detach().

ok by msaitoh@n.o
XXX pullup-8, pullup-9

fix softint_disestablish() in wm_detach().

ok by msaitoh@n.o

XXX pullup-8, pullup-9

 Add Intel I219 LM10-LM15 and V10-V14 from OpenBSD.

No longer need to go to splnet() before calling ifmedia_ioctl()
in the NET_MPSAFE; ifmedia_ioctl() will do this for us, if needed.

 KNF. No functional change.

fix build failure. ok by msaitoh@n.o.

Add SFP support part 2.

 - Add code for 1000BASE-T SFP. It uses SGMII.
 - Add extra delay in wm_serdes_power_up_link_82575(). Same as other OSes.

TODO:
 - Module insertion/removal support. Currently, SFP detection is only done
  in the driver's attach phase.

Pull up following revision(s) (requested by msaitoh in ticket #541):

	sys/dev/pci/if_wm.c: revision 1.651

 Fix a bug that wm_sgmii_readreg_locked() may return error even if error isn't
occurred.

Add SFP support part 1. The code for SerDes and SGMII setting is not include
in this commit. SFP module removal interrupt detects but not used yet:
- Detect SFP's 1000BASE-SX, 1000BASE-LX, 100BASE-FX and 1000BASE-T correctly.
- Detect the Media Auto Sense feature. Not supported yet.
- Add comment.

- Set CTRL_ILOS(Invert loss of signal) bit correctly on 82580 port 1, 2, 3 and
  newer chips. This change fixes a bug that some fiber, serdes or SFP devices
  don't detect the link status correctly.
  XXX We should check for other NVM reads if they have the same problem. One of
  the solution to avoid this type of bug is to check the register layout
  in wm_nvm_read().
- Use __BIT()
- Add comment

- MSI-X doesn't use sc->sc_icr variable, so move the code into non-MSI-X part.
  No functional change intended. OK'd by knakahara.
- Modify debug printfs a bit.

No functional change:
 - Add defintion of CONNSW register and PCS_NPTX (not used yet).
 - Add some bit definitions of PCS_LCTL.
 - Renae macro.
 - Fix comment. Add comment.
 - KNF.

 Fix a bug that wm_sgmii_readreg_locked() may return error even if error isn't
occured.

 Simplify. No functional change.

Pull up the following revisions, requested by msaitoh in ticket #1428:

	sys/dev/pci/if_wm.c			1.647, 1.649 via patch

- Use device_printf() instead of printf().
- Fix typo in comment.

Pull up following revision(s) (requested by msaitoh in ticket #404):

	sys/dev/pci/if_wm.c: revision 1.646
	sys/dev/pci/if_wm.c: revision 1.647
	sys/dev/pci/if_wm.c: revision 1.649

Use unsigned to avoid undefined behavior in wm_i82543_mii_sendbits().
Found by kUBSan.

printf -> device_printf

Fix typo in comment.

 Fix typo in comment.

 if_flags is neither int nor short. It's unsigned short.

 printf -> device_printf

Pull up following revision(s) (requested by msaitoh in ticket #1317):

	sys/dev/pci/if_wm.c: revision 1.646
	sys/dev/pci/if_wm.c: revision 1.612

Accept IFM_NONE.

 -

Use unsigned to avoid undefined behavior in wm_i82543_mii_sendbits().
Found by kUBSan.

Use unsigned to avoid undefined behavior in wm_i82543_mii_sendbits().
Found by kUBSan.

Pull up following revision(s) (requested by msaitoh in ticket #1312):

	sys/dev/pci/if_wm.c: revision 1.644
	sys/dev/pci/if_wm.c: revision 1.645
	sys/dev/pci/if_wmreg.h: revision 1.115

 Define RETA_ENTRY_MASK_Q(used in RSS redirection table) correctly.
Found by UBSan.

 Use aprint_debug_dev() instead of printf() in wm_ulp_disable().

- rnd_add_*() call must be serialized. Unserialized accesses broke the
  sc->rnd_source structure and caused off by one in kern_rndq.c's
  rnd_add_data_ts(). This problem was found by kUBSan. Make krndsource_t's
  data per queue-pair and call rnd_add_uint32() in wm_txrxintr_msix(). One of
  the reason why rnd_add_uint32() calls were moved from wm_{tx,rx}eof() is that
  rnd(9) manual page says "To allow the system to gather the timing information
  accurately, this call should be placed within the actual hardware interrupt
  service routine."

- Pass a better value to rnd_add_uint32() in wm_intr_legacy().
XXX pullup-8.

- rnd_add_*() call must be serialized. Unserialized accesses broke the
  sc->rnd_source structure and caused off by one in kern_rndq.c's
  rnd_add_data_ts(). This problem was found by kUBSan. Make krndsource_t's
  data per queue-pair and call rnd_add_uint32() in wm_txrxintr_msix(). One of
  the reason why rnd_add_uint32() calls were moved from wm_{tx,rx}eof() is that
  rnd(9) manual page says "To allow the system to gather the timing information
  accurately, this call should be placed within the actual hardware interrupt
  service routine."
- Pass a better value to rnd_add_uint32() in wm_intr_legacy().

XXX pullup-8.

 Use aprint_debug_dev() instead of printf() in wm_ulp_disable().

wm: change one bus_dmamap_load_mbuf() to bus_dmamap_load() equivalent

This makes wm work on aarch64, where the bus_dma implementation requires
the mbuf data pointer be immutable between bus_dmamap_load_mbuf and
bus_dmamap_sync. Fix suggested by thorpej@

Pull up the following revisions (via patch), requested by msaitoh in
ticket #1298:

	sys/dev/pci/if_wm.c				1.633, 1.637-1.641
	sys/dev/pci/if_wmreg.h				1.114

- Use unsigned to avoid undefined behavior.
- Print ICH/PCH's NVM version.
- Fix typo in comment. u_int*_t -> uint*_t. Whitespace fix.

 Don't automatically set ec_capenable's ETHERCAP_VLAN_HWTAGGING bit in
vlan_config() to make it user-controllable. Instead, set the bit in
xxx_attach().

 Print ICH/PCH's NVM version.

Sync with HEAD

 Use unsigned. Found by UBSan. OK'd by kamil.

s/recieve/receive/

 Whitespace fix (mainly tabify).

-No functional change:
  - KNF
  - u_int*_t -> uint*_t.

Store IFF_ALLMULTI in ec_flags instead of if_flags to avoid data races

IFF_ALLMULTI is set/unset to if_flags via if_mcast_op.  To avoid data races on
if_flags, IFNET_LOCK was added for if_mcast_op.  Unfortunately it produces
a deadlock so we want to remove added IFNET_LOCK by avoiding the data races by
another approach.

This fix introduces ec_flags to struct ethercom and stores IFF_ALLMULTI to it.
ec_flags is protected by ETHER_LOCK and thus IFNET_LOCK is no longer necessary
for if_mcast_op.  Note that the fix is applied only to MP-safe drivers that
the data races matter.

In the kernel, IFF_ALLMULTI is set by a driver and used by the driver itself.
So changing the storing place doesn't break anything.  One exception is
ioctl(SIOCGIFFLAGS); we have to include IFF_ALLMULTI in a result if needed to
export the flag as well as before.

A upcoming commit will remove IFNET_LOCK.

PR kern/54189

Pull up following revision(s) (requested by msaitoh in ticket #1267):

	sys/arch/arm/imx/if_enet.c: revision 1.19
	sys/dev/pci/if_wm.c: revision 1.634

SIOCS is 'S'et function and the ioctl argument is ifreq.
SIOCG is 'G'et function and the ioctl armument is ifmediareq.

Before this change, SIOCG modify sc->sc_flowflags unexpectedly.


Don't hook SIOCGIFMEDIA becauise this driver uses MII(4) and ether_ioctl has
the hook(if_ethersubr.c rev. 1.160).
real bug because the modification is overriden in ifmedia_ioctl().

 This driver might require some additional fixes for SIOCSIFMTU and other
ioctl()s.
XXX pullup-[78].

Remove unnecessary checks of IFF_ALLMULTI

IFF_ALLMULTI is changed by only driver itself, so we don't need to check its
change on ec_ifflags_cb.

This is part of PR kern/54189. NFCI.

SIOCS is 'S'et function and the ioctl argument is ifreq.
SIOCG is 'G'et function and the ioctl argument is ifmediareq.
Before this change, SIOCG modify sc->sc_flowflags unexpectedly.
Don't hook SIOCGIFMEDIA because this driver uses MII(4) and ether_ioctl has
the hook(if_ethersubr.c rev. 1.160). This driver might require some additional
fixes for SIOCSIFMTU and other ioctl()s.

XXX pullup-[78].

 KNF. No functional change.

 Pass error status to the upper layer in wm_gmii_{i82544,i80003}*().

Pull up the following, requested by msaitoh in ticket #1206:

	sys/dev/pci/if_wmreg.h			1.113
	sys/dev/pci/if_wmvar.h			1.43-1.44
	sys/dev/pci/if_wm.c			1.626-1.627, 1.629-1.636 via patch

- Add support for I210 SGMII Flash-less device.
- Add I219 variations for Cannon Lake.
- Add missing rnd_detach_source().
- Use  do { ... } while (0) for DPRINTF(x, y).
- Swap enumeration of 82578 and 82577 PHY in if_wmvar.h. No functional
  change.
- Add KASSERT.
- Update TODO list. Add comment. Fix typo in comment.

 Add Intel I219 variations for Cannon Lake.

 Use  do { ... } whille (0) for DPRINTF(x, y).

Typo (s/vaule/value/)

Fix GCC7 "avoid ambiguous 'else'" build error in if_wm.c

Detected with kUBSan NetBSD/i386 build.

Fix panic when wm(4) is reattached. ok by msaitoh@n.o.

This panic occured after kern_rndq.c:r1.90.

Pull up following revision(s) (requested by msaitoh in ticket #1189):

	sys/dev/pci/if_wm.c: revision 1.622
	sys/dev/pci/if_wm.c: revision 1.624
	sys/dev/pci/if_wm.c: revision 1.625

 We now have IFM_2500_KX and IFM_1000_KX, so use them.

 -

 I21[01]'s RAL size is not 15 but 16(== 82575's value). Same as other OSes.

 -

 Fix a bug that WOL didn't work on some chips since if_wm.c rev. 1.610.

Set WUC_APME bit older than PCH. Will fixes PR kern/53945 reported by kardel@.
Tested with my own 82574 card.

- Add support for I210 SGMII Flash-less device.
- Add KASSERT.
- Update TODO list.

 Fix a bug that WOL didn't work on some chips since if_wm.c rev. 1.610.
Set WUC_APME bit older than PCH. Will fixes PR kern/53945 reported by kardel@.
Tested with my own 82574 card.

 I21[01]'s RAL size is not 15 but 16(== 82575's value). Same as other OSes.

 Add yet another PHY workaround. Disable generation of early preamble on 82577
PHY rev. 1 or 2 and 82578 PHY rev. 1.

Pull up the following, requested by msaitoh in ticket #1179:

	sys/dev/pci/if_wm.c			1.603-1.605,1.607-1.611,
						1.613,1.615,1.618-1.620
						via patch
	sys/dev/pci/if_wmreg.h			1.110-1.111
	sys/dev/pci/if_wmvar.h			1.40-1.42
	sys/dev/mii/inbmphyreg.h		1.13-1.15

- Add some code for suspend/resume:
  - Rename wm_smbustopci() to wm_init_phy_workarounds_pchlan(). It will
    also called when resume.
  - Call wm_phy_resetisblocked() after PHY reset in
    wm_init_phy_workarounds_pchlan() to wait for the PHY to quiesce to
    an accessible state.
  - Add new wm_resume_workarounds_pchlan() function and use it in
    wm_resume(). This workaround is only for PCH2 and newer.
  - Don't call wm_disable_aspm() neither in wm_attach() nor in
    wm_resume() but in wm_reset().
  - Do some initialization in wm_resume() when IFF_UP is NOT set.
  - Don't continue when it failed to acquire semaphore in
    wm_ulp_disable().
- Print CLSEM workaround bit correctly.
- Fix availability detection of WoL on some chips.
- Print the WUS (WakeUp Status) register bits when resume.
- Don't setup WoL on non-WoL capable port.
- Setup PHY wakeup feature on PCH and newer. Tested on Thinkpad X220.
- Remove an extra register read in
  wm_kmrn_lock_loss_workaround_ich8lan().
- Don't leave the MDICNFG register modified when the Power Management
  capability offset can't get.
- Reduce indent level of wm_linkintr_gmii(). No functional change.
- 80003's SERDES is not the same as 82575's but the same as legacy
  devices. Use the old methods on 80003.
- Use __nothing for null DPRINTF().
- Rename functions. Add comment.

 We now have IFM_2500_KX and IFM_1000_KX, so use them.

Add some workarounds which use EMI register or EEE related:
 - PCH only: Add workaround for link disconnects on a busy hub in half duplex.
 - PCH and PCH2 only: Set MSE higher to enable link to stay up when noise is
   high.
 - PCH2 only: Drop link after 5 times MSE threshold was reached.
 - PCH2 only: Set EEE LPI Update Timer to 200usec.
 - For PCH2 and newer: When connected at 10Mbps half-duplex, some parts are
   excessively aggressive resulting in many collisions. To avoid this, increase
   the IPG and reduce Rx latency in the PHY.
 - For I21[789] and if EEE is enabled: Disable LPLU if both link partners
   support 100BaseT EEE and 100Full is advertised on both ends of the link, and
   enable Auto Enable LPI since there will be no driver to enable LPI while in
   Sx.

Sync with HEAD

 80003's SERDES is not the same as 82575's but the same as legacy devices.
Use the old methods on 80003.

XXX The reason why this bug existed is that our order of WM_T_* was little
different from FreeBSD's enum e1000_mac_type. From 80003 to PCH_CNP and from
82575 to I211 are swapped. I carefully read through if_wm.c and knew only
serdes related test were wrong. When we change WM_T_* to match FreeBSD's,
additional changes also will be required.

 Use __nothing (== ((void)0)) instead of "do {} while (/*CONSTCOND*/false)".

No functional change intended:
- Use "do {} while (/*CONSTCOND*/false)" for null DPRINTF().
- Reduce indent level of wm_linkintr_gmii().

 Change MII PHY read/write API from:

	int (*mii_readreg_t)(device_t, int, int);
	void (*mii_writereg_t)(device_t, int, int, int);
to:

	int (*mii_readreg_t)(device_t, int, int, uint16_t *);
	int (*mii_writereg_t)(device_t, int, int, uint16_t);

Now we can test if a read/write operation failed or not by the return value.

 In 802.3 spec says that the PHY shall not respond to read/write transaction
to the unimplemented register(22.2.4.3). Detecting timeout can be used to
check whether a register is implemented or not (if the register conforms to
the spec). ukphy(4) can be used this for MII_MMDACR and MII_MMDAADR.

 Note that I noticed that the following code do infinite loop in the
read/wirte function. If it accesses unimplemented PHY register, it will hang.
It should be fixed:

	arm/at91/at91emac.c
	arm/ep93xx/epe.c
	arm/omap/omapl1x_emac.c
	mips/ralink/ralink_eth.c
	arch/powerpc/booke/dev/pq3etsec.c(read)
	dev/cadence/if_cemac.c <- hkenken
	dev/ic/lan9118.c


Tested with the following device:

	axe+ukphy
	axe+rgephy
	axen+rgephy	(tested by Andrius V)
	wm+atphy
	wm+ukphy
	wm+igphy
	wm+ihphy
	wm+makphy
	sk+makphy
	sk+brgphy
	sk+gentbi
	msk+makphy
	sip+icsphy
	sip+ukphy
	re+rgephy
	bge+brgphy
	bnx+brgphy
	gsip+gphyter
	rtk+rlphy
	fxp+inphy	(tested by Andrius V)
	tlp+acphy
	ex+exphy
	epic+qsphy
	vge+ciphy	(tested by Andrius V)
	vr+ukphy	(tested by Andrius V)
	vte+ukphy	(tested by Andrius V)

Not tested (MAC):
	arm:at91emac
	arm:cemac
	arm:epe
	arm:geminigmac
	arm:enet
	arm:cpsw
	arm:emac(omac)
	arm:emac(sunxi)
	arm:npe
	evbppc:temac
	macppc:bm
	macppc:gm
	mips:aumac
	mips:ae
	mips:cnmac
	mips:reth
	mips:sbmac
	playstation2:smap
	powerpc:tsec
	powerpc:emac(ibm4xx)
	sgimips:mec
	sparc:be
	sf
	ne(ax88190, dl10019)
	awge
	ep
	gem
	hme
	smsh
	mtd
	sm
	age
	alc
	ale
	bce
	cas
	et
	jme
	lii
	nfe
	pcn
	ste
	stge
	tl
	xi
	aue
	mue
	smsc
	udav
	url

Not tested (PHY):
	amhphy
	bmtphy
	dmphy
	etphy
	glxtphy
	ikphy
	iophy
	lxtphy
	nsphyter
	pnaphy
	rdcphy
	sqphy
	tlphy
	tqphy
	urlphy

Synch with HEAD

 Set I82579_LPI_CTRL register correctly in wm_set_eee_pchlan().

- Remove an extra register read in wm_kmrn_lock_loss_workaround_ich8lan().
- Don't leave the MDICNFG register modified when the Power Management
  capability offset can't get.

 Add EEE(802.3az) support for I350, I210, I211, PCH2 and newer.

 Not yet for I354(C2000). It'll be supported after implementing MI MII clause
45 register read/write API.

 Fix error check. reported by David Binderman in kern/53821.

 Accept IFM_NONE.

Sync with HEAD, resolve a few conflicts

 Fix WoL detection once again.

- Don't setup WoL on non-WoL capable port.
- Setup PHY wakeup feature on PCH and newer. Tested on Thinkpad X220.

 Fix WoL capability detection on ICH and PCH.

 Rename functions. Add comment. No functional change.

Pull up following revision(s) (requested by knakahara in ticket #1138):

	sys/dev/pci/if_wm.c: revision 1.606

Fix txqueue assignment. Pointed out by yamaguchi@n.o, thanks.

E.g. When ncpu is six and nqueue is four, the sequence error occurs.

XXX pullup-8

 When sc->phy.acquire() failed, it's not required to call sc->phy.release().

Fix txqueue assignment. Pointed out by yamaguchi@n.o, thanks.

E.g. When ncpu is six and nqueue is four, the sequence error occurs.

XXX pullup-8

 Print and clear the wakeup status only when the WUS register != 0.

- Fix availability detection of WoL on some chips. This change has no effect
other than dmesg because WM_F_WOL is currently not used to change the behavior:
  - For ICH/PCH, check the capability not from NVM but from the WUC register.
    Check the value before clearing the register.
  - 82580 and newer have per-port NVM block, so read the area correctly.
    Note that 82580, I350 and I354 may have PCI function 2 and 3.
  - Some devices can't detect WoL capability neither from NVM nor from WUC.
    Use PCI device ID and the function number.
- Print the WUS (WakeUp Status) register bits when resume.

Add some code for suspend/resume:
- Rename wm_smbustopci() to wm_init_phy_workarounds_pchlan(). It will also
  called when resume.
- Call wm_phy_resetisblocked() after PHY reset in
  wm_init_phy_workarounds_pchlan() to wait for the PHY to quiesce to an
  accessible state.
- Add new wm_resume_workarounds_pchlan() function and use it in wm_resume().
  This workaround is only for PCH2 and newer.
- Don't call wm_disable_aspm() neither in wm_attach() nor in wm_resume()
  but in wm_reset().
- Do some initialization in wm_resume() when IFF_UP is NOT set.
- Don't continue when it failed to acquire semaphore in wm_ulp_disable().
- Add comment.

Pull up following revision(s) (requested by msaitoh in ticket #1117):

	sys/dev/pci/if_wmreg.h: revision 1.109
	sys/dev/pci/if_wm.c: revision 1.597
	sys/dev/pci/if_wm.c: revision 1.598
	sys/dev/mii/inbmphyreg.h: revision 1.12
	sys/dev/pci/if_wm.c: revision 1.600
	sys/dev/pci/if_wm.c: revision 1.601
	sys/dev/pci/if_wm.c: revision 1.602

- Add new wm_gmii_{hv,i82544}_{read,write}reg_locked() and use them in
   wm_gmii_{hv,i82544}_{read,write}reg(). *_locked() functions are not
   mii(4) API functions, so it's not required to keep the mii API. Change
   the PHY register type from int to uint16_t. It also change the usage of
   return value. It returns zero on success and non-zero on error.
- Check the return value of *_locked() function and treat it.
- Use *writereg_locked() function to reduce race condition in
   wm_init_lcd_from_nvm().
- Add comment.
- Control TX/RX descriptor snooping control bits on ICH8 and newer.
   Only on ICH8, No-snoop bits are opposite polarity. On my Thinkpad X61,
   the default value of this bits are all zero, so this commit changes
   the snoop function enable on the machine. I tested with some other
   PCH machines and those bits are all zero (enable snoop by default),
   so this commit won't affect to some machines.
- Disable relax ordering on 82546GB(Device ID 0x1099 and 0x10b5) or >= ICH8.
   Same as other OSes.
- Add wm_oem_bits_config_ich8lan() to control LPLU and GbE setting base on
   the NVM's info.
- Modify wm_enable_wakeup() to reduce difference against FreeBSD and Linux.
   This modification affects to ICH8 and newer devices. I217 Rapid Start
   Technology support have not written yet (it's TODO).
- Add wm_k1_workaround_lv() from FreeBSD. It's PCH2 specific:
     Workaround to set the K1 beacon duration for 82579 parts in 10Mbps.
     Disable K1 for 1000 and 100 speeds.
- Make wm_link_stall_workaround_hv() and move an 82578 specific code into it.
   Don't apply the workaround if BMCR_LOOP bit is set. Same as FreeBSD.
- Add comment. Modify comment.
  Fix comment. No functional change.

Pull up following revision(s) (requested by msaitoh in ticket #1656):

	sys/dev/pci/if_wm.c: revision 1.599

- Fix a bug that i82578 PHY can't detect correctly.
- Print PHY OUI and model number if a PHY's can't be identified.

XXX pullup-[78]

 Fix comment. No functional change.

Pull up following revision(s) (requested by msaitoh in ticket #1108):
	sys/dev/pci/if_wm.c: revision 1.599
- Fix a bug that i82578 PHY can't detect correctly.
- Print PHY OUI and model number if a PHY's can't be identified.

Sync with HEAD, resolve a couple of conflicts

- Add wm_k1_workaround_lv() from FreeBSD. It's PCH2 specifc:
    Workaround to set the K1 beacon duration for 82579 parts in 10Mbps.
    Disable K1 for 1000 and 100 speeds.
- Make wm_link_stall_workaround_hv() and move an 82578 specific code into it.
  Don't apply the workaround if BMCR_LOOP bit is set. Same as FreeBSD.
- Add comment. Modify comment.

- Add wm_oem_bits_config_ich8lan() to control LPLU and GbE setting base on
  the NVM's info.
- Modify wm_enable_wakeup() to reduce difference against FreeBSD and Linux.
  This modification affects to ICH8 and newer devices. I217 Rapid Start
  Technology support have not written yet (it's TODO).

- Fix a bug that i82578 PHY can't detect correctly.
- Print PHY OUI and model number if a PHY's can't be identified.

XXX pullup-[78]

- Control TX/RX descriptor snooping control bits on ICH8 and newer.
  Only on ICH8, No-snoop bits are opposite polarity. On my Thinkpas X61,
  the default value of this bits are all zero, so this commit changes
  the snoop function enable on the machine. I tested with some other
  PCH machines and those bits are all zero (enable snoop by default),
  so this commit won't affect to some machines.
- Disable relax ordering on 82546GB(Device ID 0x1099 and 0x10b5) or >= ICH8.
  Same as other OSes.

- Add new wm_gmii_{hv,i82544}_{read,write}reg_locked() and use them in
  wm_gmii_{hv,i82544}_{read,write}reg(). *_locked() functions are not
  mii(4) API functions, so it's not required to keep the mii API. Change
  the PHY register type from int to uint16_t. It also change the usage of
  return value. It returns zero on success and non-zero on error.
- Check the return value of *_locked() function and treat it.
- Use *writereg_locked() function to reduce race condition in
  wm_init_lcd_from_nvm().
- Add comment.

Pull up the following revisions(s) (requested by msaitoh in ticket #1648):
	sys/dev/mii/inbmphyreg.h		1.11
	sys/dev/pci/if_wm.c			1.586-1.588, 1.590-1.596 via patch
	sys/dev/pci/if_wmreg.h			1.108

- rename tu event counter to txunderrun.
- Try m_defrag() to reduce the number of DMA segment if bus_dmamap_load_mbuf()
   returned EFBIG. When m_defrag() is called, txqNNdefrag event counter is
   incremented. If the 2nd try of bus_dmamap_load_mbuf() failed, txqNNtoomanyseg
   event counter is incremented.
- Reduce the max number of DMA segments from 256 to 64 (it's the same value
   as other BSD's (EM_MAX_SCATTER) and more than before if_wm.c rev. 1.75's
   value (40)) because we do m_defrag() now.
- 82574 and newer document says the status field has neither EC
   (Excessive Collision) bit nor LC (Late Collision) bit (reserved), so
   don't check the bit.
- Add workaround for DMA hang problem which result in TX device timeout
   on PCH_LPT with I218. Same as FreeBSD and Linux. This workaround is
   only for device ID 0x155a, 0x15a2, 0x1559 and 0x15a3.
- Fix a PCH2 specific bug that wrong PHY register value can be read
   when boot. Same as FreeBSD and Linux.
- After writing MDIC register, don't read quickly the same register but
   do delay(50). Same as other OSes.
- Add missing wm_gate_hw_phy_config_ich8lan(false) in
   wm_phy_post_reset() on PCH2. wm_gate_hw_phy_config_ich8lan(true) is
   called in wm_reset(), so wm_gate_hw_phy_config_ich8lan(false) should
   be called after reset in wm_phy_post_reset().
- On PCH2, set the phy config counter to 50msec after (PHY) reset.
- KNF & Use macro.

Pull up the following, requested by msaitoh in tickt #1085:

	sys/dev/mii/inbmphyreg.h		1.11
	sys/dev/pci/if_wmreg.h			1.108
	sys/dev/pci/if_wm.c			1.590-1.596 via patch

- 82574 and newer document says the status field has neither EC
  (Excessive Collision) bit nor LC (Late Collision) bit (reserved), so
  don't check the bit.
- Add workaround for DMA hang problem which result in TX device timeout
  on PCH_LPT with I218. Same as FreeBSD and Linux. This workaround is
  only for device ID 0x155a, 0x15a2, 0x1559 and 0x15a3.
- Fix a PCH2 specific bug that wrong PHY register value can be read
  when boot. Same as FreeBSD and Linux.
- After writing MDIC register, don't read quickly the same register but
  do delay(50). Same as other OSes.
- Add missing wm_gate_hw_phy_config_ich8lan(false) in
  wm_phy_post_reset() on PCH2. wm_gate_hw_phy_config_ich8lan(true) is
  called in wm_reset(), so wm_gate_hw_phy_config_ich8lan(false) should
  be called after reset in wm_phy_post_reset().
- On PCH2, set the phy config counter to 50msec after (PHY) reset.
- KNF & Use macro.

make this compile again (rv was uninitialized)

 KNF a bit. No functional change.

- Add missing wm_gate_hw_phy_config_ich8lan(false) in wm_phy_post_reset()
  on PCH2. wm_gate_hw_phy_config_ich8lan(true) is called in wm_reset(), so
  wm_phy_post_reset(false) should be called after reset.
- On PCH2, set the phy config counter to 50msec after (PHY) reset.

 After writing MDIC register, don't read quickly the same register but
do delay(50). Same as otehr OSes.

 Fix a PCH2 specific bug that wrong register value can be read when boot.
When a wrong value is read when boot, the read device ID was incorrect and
ukphy(3) is attached instead of ihphy(4). The bug might also result in
MDIC read/write error.

 How to reproduce:

  0) Boot Windows.
  1) Leave some minutes.
  2) Reboot to NetBSD.

 To fix this problem, adding extra 100us delay at the end of
wm_gmii_mdic_{read,write}reg() on PCH2. Same as FreeBSD and linux.

 Reported by David Brownlee a few days ago and also reported by jmcneill
a half year ago. Tested with my own Thinkpad X220.

XXX pullup-[78]

 Add workaround for DMA hang problem which result in TX device timeout on
PCH_LPT with I218. Same as FreeBSD and Linux. This woraround is only for
device ID 0x155a, 0x15a2, 0x1559 and 0x15a3. I observed this problem on my
Shuttle DS57U's I218 port six times in two months.

- 82574 and newer's document says the status field has neither EC
  (Excessive Collision) bit nor LC (Late Collision) bit (reserved).
  Refer "PCIe GbE Controller Open Source Software Developer's Manual",
  82574 datasheet and newer.

  XXX I saw the LC bit was set on I218 even though the media was full duplex,
  so the bit might be used for other meaning ...(I have no document).
- Use macro.

Pull up the following revisions(s) (requested by msaitoh in ticket #1075):
	sys/dev/pci/if_wm.c:	revision 1.589

Fix KASSERT to prevent panic on CNP (Intel 300 series + I219) with INTx
interrupt (Xen dom0 or pre netbsd-8)

Pull up the following revisions(s) (requested by msaitoh in ticket #1647):
	sys/dev/pci/if_wm.c:	revision 1.589

Fix KASSERT to prevent panic on CNP (Intel 300 series + I219) with INTx
interrupt (Xen dom0 or pre netbsd-8)

Sync with head

 Fix KASSERT to prevent panic on CNP(Intel 300 series + I219) with INTx
interrupt (Xen dom0 or pre netbsd-8).

XXX pullup-[78]

Ssync with HEAD

Pull up following revision(s) (requested by msaitoh in ticket #1027):
	sys/dev/pci/if_wm.c: revision 1.586
	sys/dev/pci/if_wm.c: revision 1.587
	sys/dev/pci/if_wm.c: revision 1.588

- Split txdrop evcnt into pcqdrop and descdrop.
- Simplify evcnt name by removing duplicated "tx" or "rx".
- s/tu/underrun/

- Try m_defrag() to reduce the number of DMA segment if bus_dmamap_load_mbuf()
   returned EFBIG. When m_defrag() is called, txqNNdefrag event counter is
   incremented. If the 2nd try of bus_dmamap_load_mbuf() failed, txqNNtoomanyseg
   event counter is incremented.

- Reduce the max number of DMA segments from 256 to 64 (it's the same value
   as other BSD's (EM_MAX_SCATTER) and more than before if_wm.c rev. 1.75's
   value (40)) because we do m_defrag() now.

  Fix comment.

 Fix comment.

- Try m_defrag() to reduce the number of DMA segment if bus_dmamap_load_mbuf()
  returned EFBIG. When m_defrag() is called, txqNNdefrag event counter is
  incremented. If the 2nd try of bus_dmamap_load_mbuf() failed, txqNNtoomanyseg
  event counter is incremented.

- Reduce the max number of DMA segemts from 256 to 64 (it's the same valuse
  as other BSD's (EM_MAX_SCATTER) and more than before if_wm.c rev. 1.75's
  value (40)) because we do m_defrag() now.

- Split txdrop evcnt into pcqdrop and descdrop.
- Simplify evcnt name by removing duplicated "tx" or "rx".
- s/tu/underrun/

Sync with HEAD

Resolve a couple of conflicts (result of the uimin/uimax changes)

Rename min/max -> uimin/uimax for better honesty.

These functions are defined on unsigned int.  The generic name
min/max should not silently truncate to 32 bits on 64-bit systems.
This is purely a name change -- no functional change intended.

HOWEVER!  Some subsystems have

	#define min(a, b)	((a) < (b) ? (a) : (b))
	#define max(a, b)	((a) > (b) ? (a) : (b))

even though our standard name for that is MIN/MAX.  Although these
may invite multiple evaluation bugs, these do _not_ cause integer
truncation.

To avoid `fixing' these cases, I first changed the name in libkern,
and then compile-tested every file where min/max occurred in order to
confirm that it failed -- and thus confirm that nothing shadowed
min/max -- before changing it.

I have left a handful of bootloaders that are too annoying to
compile-test, and some dead code:

cobalt ews4800mips hp300 hppa ia64 luna68k vax
acorn32/if_ie.c (not included in any kernels)
macppc/if_gm.c (superseded by gem(4))

It should be easy to fix the fallout once identified -- this way of
doing things fails safe, and the goal here, after all, is to _avoid_
silent integer truncations, not introduce them.

Maybe one day we can reintroduce min/max as type-generic things that
never silently truncate.  But we should avoid doing that for a while,
so that existing code has a chance to be detected by the compiler for
conversion to uimin/uimax without changing the semantics until we can
properly audit it all.  (Who knows, maybe in some cases integer
truncation is actually intended!)

Pull up following revision(s) (requested by msaitoh in ticket #969):

	sys/dev/pci/if_wm.c: revision 1.584
	sys/dev/pci/if_wmvar.h: revision 1.39

Add MDIC_WA and CLSEMWA to snprintb() string for WM_F_* flags.

Fix two bugs for non-SERDES fiber. Tested on 82544EI(0x1009), 82545GM(0x1027),
82546GB(0x107a), 82571GB(0x105f), 82572EI(0x107e). Tested on SERDES
82580(0x150e (Winyao SERDES SFP)) and 82580(0x150e) as regression test:

- Fix a bug that the link can't detect in link interrupt function for
   non-SERDES fiber. When the signal is detected & STATUS_LU == 0 & RXCW == 0,
   it's required to set CTRL_SLU and CTRL_FD bit. wm_check_for_link() is the
   function to do it. If a link status change interrupt occurred, call
   wm_check_for_link() before checking current link status.

- Fix a bug that 82542 misunderstand fiber's signal detection. CTRL_SWDPIN(1)
   is used for fiber's signal and it's value is reversed only on 82543 and
   82544. Simplify the code and move it into a new function wm_tbi_havesignal()
   and use it.

Pull up the following, requested by msaitoh in ticket #1628:

share/man/man4/wm.4				1.40 via patch
sys/dev/mii/ihphyreg.h				1.2
sys/dev/mii/inbmphyreg.h			1.10
sys/dev/pci/if_wm.c				1.504, 1.506, 1.510-1.535, 1.539-1.540, 1.546, 1.548, 1.551-1.552, 1.558, 1.565-1.573, 1.575, 1.579, 1.582, 1.584 via patch
sys/dev/pci/if_wmreg.h				1.99-1.103, 1.106-1.107 via patch
sys/dev/pci/if_wmvar.h				1.34-1.39 via patch
sys/dev/pci/pcidevs				1.1327 via patch
sys/dev/pci/pcidevs.h				regen
sys/dev/pci/pcidevs_data.h			regen
sys/dev/pci/pcireg.h				patch

	Sync wm(4) up to 2018/08/08 except MSI/MSI-X and NET_MPSAFE:
	- remove extra "+"
	- Fix a bug that non-GMII devices don't send a routing message when
	  the link status is changed.
	- Set WMREG_KABGTXD not in wm_init_locked() but in wm_reset(). Same as
	  other OSes.
	- If a interrupt is a spurious interrupt, don't print debug message.
	- Don't print the Image Unique ID if an NVM is iNVM (i210 and I211).
	- Print sc_flags with snprintb().
	- Fix a bug that a RAL was written at incorrect address when the index
	  number is more than 16 on 82544 and newer.
	- The layout of RAL on PCH* are different from others. Fix it.
	- Flush every MTA write. Same as Linux.
	- Move the location of calling wm_set_filter. Same as some other OSes.
	- Add CSR_WRITE_FLUSH() after writing WMREG_CTRL in
	  wm_gmii_mediachange().
	- Add missing "else" in wm_nvm_release().
	- Make new wm_phy_post_reset() and use this function at all location
	  after resetting phy.
	- Move the location of calling wm_get_hw_control. Same as Linux.
	- Add I219 specific wokaround for legacy interrupt. From OpenBSD.
	- Move the location of calling wm_lplu_d0_disable().
	- Fix latency calculation in wm_platform_pm_pch_lpt().
	- Set OBFF water mark and enable OBFF on PCH_LPT and newer.
	- Disable D0 LPLU on 8257[12356], 82580, I350 and I21[01], too.	Before
	  this commit, above devices and non-PCIe devices accessed wrong
	  register.
	- Use device_printf() instead of aprint_error_dev() for PHY read/write
	  functions because those are used not only in device attach.
	- Fix a bug that wm_gmii_i82544_{read,write}reg() didn't take care of
	  page select. PHY access from igphy() automatically did it, but
	  accessing from wm(4) for wrokaround didn't work correctly. This
	  change affects 8254[17], 8257[12] ICH8, ICH9 and ICH10.
	- Call wm_kmrn_lock_loss_workaround_ich8lan() before any PHY access in
	  wm_linkintr_gmii().
	- Register access in wm_kmrn_lock_loss_workaround_ich8lan() now works
	  correctly. Enable this function.
	- Configure the LCD with the extended configuration region in NVM if
	  it's required.
	- If TX is not required to flush, RX is also not required to flush
	  in wm_flush_desc_rings(). Same as other OSes.
	- Remove wrong semaphore access in wm_nvm_{read,write}_{ich8,spt} to
	  prevent hangup. A semaphore is get/put in wm_nvm_{read,write}.
	- Move some initialization stuff in wm_attach() before wm_reset(). Some
	  flags and callback function are required to set correctly before
	  wm_reset() because wm_reset() and some helper functions refer them.
	- Add wm_write_smbus_addr() to set SMBus address by software.
	- Modify wm_gmii_hv_{read,write}reg_locked() to make them access
	  HV_SMB_ADDR correctly.
	- Use new nvm.{acquire,release}() for semaphore.
	- Our MII readreg/writereg API has not way to detect an error.
	  kmrn_{read,write}reg() are not used for MII API, so it's not required
	  for these functions to use the same API. So,
	  - Change return value as error code.
	  - Change register value from int to uint16_t.
	  - read: pass pointer for uint16_t as an argument.
	  - Check return value on caller side.
	- Check whether it's required to use MDIC workaround for 80003 or not
	  in wm_reset(). If the workaround isn't required, don't use the
	  workaround code in wm_gmii_i80003_{read,write}reg.
	- Add WM_F_WA_I210_CLSEM flag for a workaround. FreeBSD/Linux drivers
	  say "In rare circumstances, the SW semaphore may already be held
	  unintentionally on I21[01]". PXE boot is one of the case.
	- Qemu's e1000e emulation (82574L)'s SPI has only 64 words. I've never
	  seen on real 82574 hardware with such small SPI ROM. Check
	  sc->sc_nvm_wordsize before accessing higher address words to prevent
	  timeout.
	- Check some wm_nvm_read()'s return vale.
	- Print NVM offset and word count when EERD polling failed.
	- On I219, drop TARC0 bit 28 for DMA hang workaround (from Linux).
	- 82583 supports jumbo frame. Fixes PR#52773 reported by
	  Shinichi Doyashiki.
	- Fix typo in comment. Reported by Shinichi Doyashiki in PR#52885.
	- Add ASPM workaround for 8257[1234] and 82583 to prevent device
	  timeout or hangup. Fixes PR#52818 reported by Shinichi Doyashiki.
	- CID-1427779: Fix uninitialized variables.
	- Fix a bug that wm_pll_workaround_i210() is not called when
	  a) Chip is I211 or b) Chip is I210 and it uses iNVM (not FLASH).
	- Do wm_reset_mdicnfg_82580() on 82580 only.
	- Fix FLASH access on PCH_SPT and newer. Their FLASH access should be
	  done by 32bit. Especially for ICH_FLASH_HSFCTL register, it's located
	  at 0x0006, so it must be accessed via ICH_FLASH_HSFSTS(0x0004) and
	  use shift or mask.
	- Make wm_nvm_valid_bank_detect_ich8lan() the same as other OSes.
	- If the extended configuration size in the EXTCNFSIZE register is 0,
	  don't continue in wm_init_lcd_from_nvm().
	- Add PCH_CNP support (I219 with Intel 300 series chipset).
	- Enable I219 support.
	- I354 uses an external PHY, so don't use wm_set_eee_i350().
	- Fix a bug that the link can't detect in link interrupt function for
	  non-SERDES fiber.
	- Fix a bug that 82542 misunderstand fiber's signal detection.
	- Add debug printf()s.
	- Update comment.
	- Rename functions and variables.
	- Add diagnostic code.
	- Sort registers.
	- Lowercase hexadecimal values.
	- KNF.

Fix two bugs for non-SERDES fiber. Tested on 82544EI(0x1009), 82545GM(0x1027),
82546GB(0x107a), 82571GB(0x105f), 82572EI(0x107e). Tested on SERDES
82580(0x150e (Winyao SERDES SFP)) and 82580(0x150e) as regression test:

- Fix a bug that the link can't detect in link interrupt function for
  non-SERDES fiber. When the signal is detected & STATUS_LU == 0 & RXCW == 0,
  it's required to set CTRL_SLU and CTRL_FD bit. wm_check_for_link() is the
  function to do it. If a link status change interrupt occured, call
  wm_check_for_link() before checking current link staus.
- Fix a bug that 82542 misunderstand fiber's signal detection. CTRL_SWDPIN(1)
  is used for fiber's signal and it's value is reversed only on 82543 and
  82544. Simplify the code and move it into a new function wm_tbi_havesignal()
  and use it.

Sync with HEAD

Pull up following revision(s) (requested by msaitoh in ticket #928):
	sys/dev/pci/if_wm.c: 1.573-1.576, 1.579 and 1.582 via patch
wm(4):
- Count timeout correctly. This change reduce timeout value for 80003
  as expected. Reported by mouse@.
- Print "device timeout (lost interrupt)"  when all descriptors in a
  queue are free.
- Rename txq_watchdog to txq_sending to make the meaning clear.
- KNF. No functional change.

 Implement the BPF direction filter (BIOC[GS]DIRECTION). It provides backward
compatibility with BIOC[GS]SEESENT ioctl. The userland interface is the same
as FreeBSD.

 This change also fixes a bug that the direction is misunderstand on some
environment by passing the direction to bpf_mtap*() instead of checking
m->m_pkthdr.rcvif.

Sync with HEAD

 Style fix. No functional change.

Pull up following revision(s) (requested by ozaki-r in ticket #843):

	sys/dev/pci/ixgbe/ixv.c: revision 1.101
	sys/net/if_bridge.c: revision 1.156
	sys/net/if_pppoe.c: revision 1.138
	sys/dev/pci/if_wm.c: revision 1.580
	sys/dev/pci/ixgbe/ixgbe.c: revision 1.156
	sys/net/if_gif.c: revision 1.142

Ensure to call if_register after interface initializations finish

Rename

	M_CSUM_DATA_IPv6_HL -> M_CSUM_DATA_IPv6_IPHL
	M_CSUM_DATA_IPv6_HL_SET -> M_CSUM_DATA_IPv6_SET

Reduces the diff against IPv4. Also, clarify the definitions.

Ensure to call if_register after interface initializations finish

Sync with HEAD

Pull up following revision(s) (requested by msaitoh in ticket #827):

	sys/dev/pci/if_wm.c: revision 1.577

rxipsum and rxtusum are not interrupt counter, so use EVCNT_TYPE_MISC
instead of EVCNT_TYPE_INTR.

 KNF. No binary change.

Pull up following revision(s) (requested by msaitoh in ticket #816):

	sys/dev/pci/if_wm.c: revision 1.578

  Fix a bug that TX might stall because WM_TXQ_NO_SPACE is not cleared in
if_init() (though I've never seen this problem). Clear txq->txq_flags in
wm_init_tx_queue(). OK'd by knakahara.

 Fix a bug that TX might stall because WM_TXQ_NO_SPACE is not cleared in
if_init() (though I've never seen this problem). Clear txq->txq_flags in
wm_init_tx_queue(). OK'd by knakahara.

 rxipsum and rxtusum are not interrupt counter, so use EVCNT_TYPE_MISC
instead of EVCNT_TYPE_INTR.

Synch with HEAD

- Backout if_wm.c rev.1.574 and print "device timeout (lost interrupt)"
 when all descriptors in a queue are free. When all descriptors are free after
 wm_txeof(), it's caused by lost interrupt (though I've never seen it). One
 possibility is chip bug and another possibility is software bug. We should
 reset in any cases. If we don't reset and don't print error message, TX
 processing is done intermittently and user might not noticed the problem.

- Rename txq_watchdog to txq_sending to make the meaning clear.

 Count timeout correctly. This change reduce timeout value for 80003 as
expected. Reported by mouse@.

Sync with HEAD

 Fix watchdog timer. Without this change, watchdog timer is unnecessary
fired and device is initialized without any error message. OK'd by knakahara.

Pull up following revision(s) (requested by msaitoh in ticket #764):

	sys/dev/pci/if_wm.c: revision 1.567
	sys/dev/pci/if_wm.c: revision 1.568
	sys/dev/pci/if_wm.c: revision 1.569
	sys/dev/pci/if_wmvar.h: revision 1.38
	sys/dev/pci/if_wm.c: revision 1.570
	sys/dev/pci/if_wm.c: revision 1.571
	sys/dev/pci/if_wm.c: revision 1.572
	share/man/man4/wm.4: revision 1.40
	sys/dev/pci/if_wmreg.h: revision 1.106
	sys/dev/pci/if_wmreg.h: revision 1.107

SW PHY Config Enable bit for ICH8 B0 stepping is not bit 1 but bit 0.

 No binary change:
- Sort registers.
- Lowercase hexadecimal value.

 On PCH_SPT (and newer), FLASH access should be done by 32bit.
Especially for ICH_FLASH_HSFCTL register, it's located at 0x0006, so
it should be accessed via ICH_FLASH_HSFSTS(0x0004) and use shift or mask.

 Our PCH_SPT part of wm_nvm_valid_bank_detect_ich8lan() was based on
FreeBSD r287467. After that, they reverted it and committed the different
code in r287762. r287762's bank detect code didn't work for us because our wm
dirver had a problem in flash access. The problem was fixed in if_wm.c rev.
1.567, so we can use the new way now.

 If the extended configration size in the EXTCNFSIZE register is 0, don't
continue.
 Add PCH_CNP support (I219 with Intel 300 series chipset).
It's required more test, so it's disabled by default.

 Enable I219.

 I354 uses an external PHY, so don't use wm_set_eee_i350().

 KNF. No functional change.

Sync with HEAD, resolve some conflicts

 I354 uses an external PHY, so don't use wm_set_eee_i350().

 Enable I219.

 Add PCH_CNP support (I219 with Intel 300 series chipset).
It's required more test, so it's disabled by default.

 If the extended configration size in the EXTCNFSIZE register is 0, don't
continue.

 Our PCH_SPT part of wm_nvm_valid_bank_detect_ich8lan() was based on
FreeBSD r287467. After that, they reverted it and committed the different
code in r287762. r287762's bank detect code didn't work for us because our wm
dirver had a problem in flash access. The problem was fixed in if_wm.c rev.
1.567, so we can use the new way now.

 On PCH_SPT (and newer), FLASH access should be done by 32bit.
Especially for ICH_FLASH_HSFCTL register, it's located at 0x0006, so
it should be accessed via ICH_FLASH_HSFSTS(0x0004) and use shift or mask.

Pull up following revision(s) (requested by msaitoh in ticket #604):
	sys/dev/pci/if_wm.c: revision 1.565-1.566

- Fix a bug that wm_pll_workaround_i210() is not called on some cases.
   As I wrote in if_wm.c rev 1.329's commit message, There are three cases:
     a) Chip is I211
     b) Chip is I210 and it uses INVM (not FLASH)
     c) Chip is I210 (it uses FLASH) and NVM image version < 3.25
   Before this commit, the workaround function was not called on a) and b).

- Rename function.

Do wm_reset_mdicnfg_82580() on 82580 only.

 Do wm_reset_mdicnfg_82580() on 82580 only.

- Fix a bug that wm_pll_workaround_i210() is not called on some cases.
  As I wrote in if_wm.c rev 1.329's commit message, There are three cases:

    a) Chip is I211
    b) Chip is I210 and it uses INVM (not FLASH)
    c) Chip is I210 (it uses FLASH) and NVM image version < 3.25

  Before this commit, the workaround function was not called on a) and b).

- Rename function.

Pull up following revision(s) (requested by knakahara in ticket #567):
	distrib/sets/lists/comp/mi: 1.2182-1.2183
	sys/dev/pci/if_wm.c: 1.564
	sys/dev/pci/ixgbe/ixgbe.c: 1.122
	sys/dev/pci/ixgbe/ixgbe_rss.h: 1.3
	sys/dev/pci/ixgbe/ixv.c: 1.78
	sys/net/Makefile: 1.35-1.36
	sys/net/files.net: 1.15
	sys/net/rss_config.c: 1.1
	sys/net/rss_config.h: 1.1
Introduce very simple Receive Side Scaling (RSS) utility.
ok by msaitoh@n.o.
--
Apply RSS utility to wm(4).
ok by msaitoh@n.o.
--
Apply RSS utility to ixg(4) and ixv(4).
ok by msaitoh@n.o.
--
Fix build failure, sorry.
--
Currently, it is not necessary to install rss_config.h. Pointed out by msaitoh@n.o.

Pull up following revision(s) (requested by knakahara in ticket #564):
	sys/dev/pci/if_wm.c: 1.563
Fix a bug that RX and TX may stall on heavy load on wm(4) like ixgbe.c:r1.121.
wm_rxeof() and wm_txeof() have loop limit and the function returns true
if a packet still exists.

Apply RSS utility to wm(4).

ok by msaitoh@n.o.

Fix a bug that RX and TX may stall on heavy load on wm(4) like ixgbe.c:r1.121.

wm_rxeof() and wm_txeof() have loop limit and the function returns true
if a packet still exists.

XXX need pullup-8

Pull up following revision(s) (requested by knakahara in ticket #529):
	sys/dev/pci/if_wm.c: revision 1.560
	sys/dev/pci/if_wm.c: revision 1.561
	sys/dev/pci/if_wm.c: revision 1.562
	share/man/man4/wm.4: revision 1.37
	share/man/man4/wm.4: revision 1.38
	sys/dev/pci/if_wm.c: revision 1.551
	sys/dev/pci/if_wm.c: revision 1.553
	sys/dev/pci/if_wm.c: revision 1.554
	sys/dev/pci/if_wm.c: revision 1.555
	sys/dev/pci/if_wm.c: revision 1.556
	sys/dev/pci/if_wm.c: revision 1.557
	sys/dev/pci/if_wm.c: revision 1.558
	sys/dev/pci/if_wm.c: revision 1.559
PR/52885 - Shinichi Doyashiki -- typo in comment
Fix legacy Tx descriptors printing when WM_DEBUG is enabled.
improve comments
Fix wm_watchdog_txq() lock region.
Not only wm_txeof() but also wm_watchdog_txq() itself requires txq_lock
as it reads Tx descriptor management variables such as "txq_free".
There is almost no influence on performance.
Fix duplicated "rxintr" evcnt counting. Pointed out by ozaki-r@n.o, thanks.
wm_txeof() can limit the loop count the same as wm_rxeof() now.
add WM_TX_PROCESS_LIMIT_DEFAULT and WM_TX_INTR_PROCESS_LIMIT_DEFAULT man.
More markup.
CID-1427779: Fix uninitialized variables
Fix 82574 MSI-X mode cannot receive packets after 82574 receives high rate traffic.
In short, 82574 MSI-X mode does not cause RXQ MSI-X vector when 82574's
phys FIFO overflows. I don't know why but 82574 causes not RXQ MSI-X vector
but OTHER MSI-X vector at the situation.
see:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=v4.15-rc9&id=4aea7a5c5e940c1723add439f4088844cd26196d
advised by msaitoh@n.o, thanks.
Fix if_wm.c:r1.557 merge miss, sorry.
Fix unmatched return type. The return value of wm_txeof() is not useded yet.
Make wm(4) watchdog MP-safe. There is almost no influence on performance.
wm(4) does not use ifp->if_watchdog now, that is, it does not touch
ifp->if_timer.
It also uses own callout(wm_tick) as watchdog now. The watchdog uses
per-queue counter to check timeout. So, global lock is not required.

Make wm(4) watchdog MP-safe. There is almost no influence on performance.

wm(4) does not use ifp->if_watchdog now, that is, it does not touch
ifp->if_timer.
It also uses own callout(wm_tick) as watchdog now. The watchdog uses
per-queue counter to check timeout. So, global lock is not required.

Fix unmatched return type. The return value of wm_txeof() is not useded yet.

Fix if_wm.c:r1.557 merge miss, sorry.

Fix 82574 MSI-X mode cannot receive packets after 82574 receives high rate traffic.

In short, 82574 MSI-X mode does not cause RXQ MSI-X vector when 82574's
phys FIFO overflows. I don't know why but 82574 causes not RXQ MSI-X vector
but OTHER MSI-X vector at the situation.
see:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=v4.15-rc9&id=4aea7a5c5e940c1723add439f4088844cd26196d

advised by msaitoh@n.o, thanks.

CID-1427779: Fix uninitialized variables

wm_txeof() can limit the loop count the same as wm_rxeof() now.

Fix duplicated "rxintr" evcnt counting. Pointed out by ozaki-r@n.o, thanks.

Fix wm_watchdog_txq() lock region.

Not only wm_txeof() but also wm_watchdog_txq() itself requires txq_lock
as it reads Tx descriptor management variables such as "txq_free".

There is almost no influence on performance.

improve comments

Fix legacy Tx descriptors printing when WM_DEBUG is enabled.

Pull up following revision(s) (requested by msaitoh in ticket #491):
	sys/dev/pci/if_wm.c: 1.550, 1.552
 Don't use MSI-X if we can use only one queue to save interrupt resource.
Written by knakahara and tested by me.
--
 Add ASPM workaround for 8257[1234] and 82583 to prevent device timeout or
hangup. Fixes PR#52818 reported by Shinichi Doyashiki.

 Add ASPM workaround for 8257[1234] and 82583 to prevent device timeout or
hangup. Fixes PR#52818 reported by Shinichi Doyashiki.

Pull up following revision(s) (requested by ozaki-r in ticket #456):
	sys/arch/arm/sunxi/sunxi_emac.c: 1.9
	sys/dev/ic/dwc_gmac.c: 1.43-1.44
	sys/dev/pci/if_iwm.c: 1.75
	sys/dev/pci/if_wm.c: 1.543
	sys/dev/pci/ixgbe/ixgbe.c: 1.112
	sys/dev/pci/ixgbe/ixv.c: 1.74
	sys/kern/sys_socket.c: 1.75
	sys/net/agr/if_agr.c: 1.43
	sys/net/bpf.c: 1.219
	sys/net/if.c: 1.397, 1.399, 1.401-1.403, 1.406-1.410, 1.412-1.416
	sys/net/if.h: 1.242-1.247, 1.250, 1.252-1.257
	sys/net/if_bridge.c: 1.140 via patch, 1.142-1.146
	sys/net/if_etherip.c: 1.40
	sys/net/if_ethersubr.c: 1.243, 1.246
	sys/net/if_faith.c: 1.57
	sys/net/if_gif.c: 1.132
	sys/net/if_l2tp.c: 1.15, 1.17
	sys/net/if_loop.c: 1.98-1.101
	sys/net/if_media.c: 1.35
	sys/net/if_pppoe.c: 1.131-1.132
	sys/net/if_spppsubr.c: 1.176-1.177
	sys/net/if_tun.c: 1.142
	sys/net/if_vlan.c: 1.107, 1.109, 1.114-1.121
	sys/net/npf/npf_ifaddr.c: 1.3
	sys/net/npf/npf_os.c: 1.8-1.9
	sys/net/rtsock.c: 1.230
	sys/netcan/if_canloop.c: 1.3-1.5
	sys/netinet/if_arp.c: 1.255
	sys/netinet/igmp.c: 1.65
	sys/netinet/in.c: 1.210-1.211
	sys/netinet/in_pcb.c: 1.180
	sys/netinet/ip_carp.c: 1.92, 1.94
	sys/netinet/ip_flow.c: 1.81
	sys/netinet/ip_input.c: 1.362
	sys/netinet/ip_mroute.c: 1.147
	sys/netinet/ip_output.c: 1.283, 1.285, 1.287
	sys/netinet6/frag6.c: 1.61
	sys/netinet6/in6.c: 1.251, 1.255
	sys/netinet6/in6_pcb.c: 1.162
	sys/netinet6/ip6_flow.c: 1.35
	sys/netinet6/ip6_input.c: 1.183
	sys/netinet6/ip6_output.c: 1.196
	sys/netinet6/mld6.c: 1.90
	sys/netinet6/nd6.c: 1.239-1.240
	sys/netinet6/nd6_nbr.c: 1.139
	sys/netinet6/nd6_rtr.c: 1.136
	sys/netipsec/ipsec_output.c: 1.65
	sys/rump/net/lib/libnetinet/netinet_component.c: 1.9-1.10
kmem_intr_free kmem_intr_[z]alloced memory
the underlying pools are the same but api-wise those should match
Unify IFEF_*_MPSAFE into IFEF_MPSAFE
There are already two flags for if_output and if_start, however, it seems such
MPSAFE flags are eventually needed for all if_XXX operations. Having discrete
flags for each operation is wasteful of if_extflags bits. So let's unify
the flags into one: IFEF_MPSAFE.
Fortunately IFEF_*_MPSAFE flags have never been included in any releases, so
we can change them without breaking backward compatibility of the releases
(though the kernel version of -current should be bumped).
Note that if an interface have both MP-safe and non-MP-safe operations at a
time, we have to set the IFEF_MPSAFE flag and let callees of non-MP-safe
opeartions take the kernel lock.
Proposed on tech-kern@ and tech-net@
Provide macros for softnet_lock and KERNEL_LOCK hiding NET_MPSAFE switch
It reduces C&P codes such as "#ifndef NET_MPSAFE KERNEL_LOCK(1, NULL); ..."
scattered all over the source code and makes it easy to identify remaining
KERNEL_LOCK and/or softnet_lock that are held even if NET_MPSAFE.
No functional change
Hold KERNEL_LOCK on if_ioctl selectively based on IFEF_MPSAFE
If IFEF_MPSAFE is set, hold the lock and otherwise don't hold.
This change requires additions of KERNEL_LOCK to subsequence functions from
if_ioctl such as ifmedia_ioctl and ifioctl_common to protect non-MP-safe
components.
Proposed on tech-kern@ and tech-net@
Ensure to hold if_ioctl_lock when calling if_flags_set
Fix locking against myself on ifpromisc
vlan_unconfig_locked could be called with holding if_ioctl_lock.
Ensure to not turn on IFF_RUNNING of an interface until its initialization completes
And ensure to turn off it before destruction as per IFF_RUNNING's description
"resource allocated". (The description is a bit doubtful though, I believe the
change is still proper.)
Ensure to hold if_ioctl_lock on if_up and if_down
One exception for if_down is if_detach; in the case the lock isn't needed
because it's guaranteed that no other one can access ifp at that point.
Make if_link_queue MP-safe if IFEF_MPSAFE
if_link_queue is a queue to store events of link state changes, which is
used to pass events from (typically) an interrupt handler to
if_link_state_change softint. The queue was protected by KERNEL_LOCK so far,
but if IFEF_MPSAFE is enabled, it becomes unsafe because (perhaps) an interrupt
handler of an interface with IFEF_MPSAFE doesn't take KERNEL_LOCK. Protect it
by a spin mutex.
Additionally with this change KERNEL_LOCK of if_link_state_change softint is
omitted if NET_MPSAFE is enabled.
Note that the spin mutex is now ifp->if_snd.ifq_lock as well as the case of
if_timer (see the comment).
Use IFADDR_WRITER_FOREACH instead of IFADDR_READER_FOREACH
At that point no other one modifies the list so IFADDR_READER_FOREACH
is unnecessary. Use of IFADDR_READER_FOREACH is harmless in general though,
if we try to detect contract violations of pserialize, using it violates
the contract. So avoid using it makes life easy.
Ensure to call if_addr_init with holding if_ioctl_lock
Get rid of outdated comments
Fix build of kernels without ether
By throwing out if_enable_vlan_mtu and if_disable_vlan_mtu that
created a unnecessary dependency from if.c to if_ethersubr.c.
PR kern/52790
Rename IFNET_LOCK to IFNET_GLOBAL_LOCK
IFNET_LOCK will be used in another lock, if_ioctl_lock (might be renamed then).
Wrap if_ioctl_lock with IFNET_* macros (NFC)
Also if_ioctl_lock perhaps needs to be renamed to something because it's now
not just for ioctl...
Reorder some destruction routines in if_detach
- Destroy if_ioctl_lock at the end of the if_detach because it's used in various
  destruction routines
- Move psref_target_destroy after pr_purgeif because we want to use psref in
  pr_purgeif (otherwise destruction procedures can be tricky)
Ensure to call if_mcast_op with holding IFNET_LOCK
Note that CARP doesn't deal with IFNET_LOCK yet.
Remove IFNET_GLOBAL_LOCK where it's unnecessary because IFNET_LOCK is held
Describe which lock is used to protect each member variable of struct ifnet
Requested by skrll@
Write a guideline for converting an interface to IFEF_MPSAFE
Requested by skrll@
Note that IFNET_LOCK must not be held in softint
Don't set IFEF_MPSAFE unless NET_MPSAFE at this point
Because recent investigations show that interfaces with IFEF_MPSAFE need to
follow additional restrictions to work with the flag safely. We should enable it
on an interface by default only if the interface surely satisfies the
restrictions, which are described in if.h.
Note that enabling IFEF_MPSAFE solely gains a few benefit on performance because
the network stack is still serialized by the big kernel locks by default.

PR/52885 - Shinichi Doyashiki -- typo in comment

 Don't use MSI-X if we can use only one queue to save interrupt resource.
Written by knakahara and tested by me.

Pull up following revision(s) (requested by msaitoh in ticket #432):
	sys/dev/pci/if_wm.c: revision 1.548
Fix a bug that 8257[56] can't receive packet reported by Bert Kiers in
PR#52717. For 82575 and 82576, the RX descriptors must be initialized after
the setting of RCTL.EN in wm_set_filter(). This bug was added in if_wm.c
rev. 1.515.

Pull up following revision(s) (requested by msaitoh in ticket #427):
	sys/arch/amiga/dev/if_bah_zbus.c: 1.17
	sys/arch/arm/broadcom/bcm53xx_eth.c: 1.30
	sys/arch/powerpc/booke/dev/pq3etsec.c: 1.32
	sys/arch/usermode/dev/if_veth.c: 1.9
	sys/dev/ic/an.c: 1.66
	sys/dev/ic/athn.c: 1.17
	sys/dev/ic/atw.c: 1.162
	sys/dev/ic/bwi.c: 1.33
	sys/dev/ic/dwc_gmac.c: 1.41-1.42
	sys/dev/ic/malo.c: 1.10
	sys/dev/ic/rt2560.c: 1.31
	sys/dev/ic/rt2661.c: 1.36
	sys/dev/ic/rt2860.c: 1.29
	sys/dev/ic/rtw.c: 1.127
	sys/dev/ic/rtwvar.h: 1.46
	sys/dev/ic/smc90cx6.c: 1.71
	sys/dev/ic/smc90cx6var.h: 1.12
	sys/dev/ic/wi.c: 1.244
	sys/dev/pci/if_ipw.c: 1.66
	sys/dev/pci/if_iwi.c: 1.104
	sys/dev/pci/if_iwm.c: 1.76
	sys/dev/pci/if_iwn.c: 1.86
	sys/dev/pci/if_rtwn.c: 1.13
	sys/dev/pci/if_wm.c: 1.541
	sys/dev/pci/if_wpi.c: 1.79
	sys/dev/pci/ixgbe/ixgbe.c: 1.106
	sys/dev/pci/ixgbe/ixv.c: 1.73 via patch
	sys/dev/pcmcia/if_malo_pcmcia.c: 1.15
	sys/dev/scsipi/if_se.c: 1.95
	sys/dev/usb/if_upl.c: 1.60
	sys/net/if.c: 1.396
	sys/net/if.h: 1.241
	sys/net/if_arc.h: 1.23
	sys/net/if_arcsubr.c: 1.78
	sys/net/if_bridge.c: 1.136-1.137
	sys/net/if_etherip.c: 1.39
	sys/net/if_faith.c: 1.56
	sys/net/if_gif.c: 1.131
	sys/net/if_loop.c: 1.96
	sys/net/if_mpls.c: 1.30
	sys/net/if_pppoe.c: 1.129
	sys/net/if_srt.c: 1.27
	sys/net/if_stf.c: 1.102
	sys/net/if_tap.c: 1.100
	sys/net/if_vlan.c: 1.105
	sys/netinet/ip_carp.c: 1.91
	sys/rump/net/lib/libshmif/if_shmem.c: 1.73-1.74
	sys/rump/net/lib/libvirtif/if_virt.c: 1.55-1.56
 if_initalize() and if_attach() failed when resource allocation failed
(e.g. allocating softint). Without this change, it panics. It's bad because
resource shortage really occured when a lot of pseudo interface is created.
To avoid this problem, don't panic and change return value of if_initialize()
and if_attach() to int. Caller fanction will be recover from error cleanly by
checking the return value.
 Return if bah_attach_subr() failed.
 If if_attach() failed in the attach function, return.
- If if_initialize() failed in the attach function, free resources and return.
- Add some missing frees in bridge_clone_destroy().
- KNF
 If error occured in bcmeth_ccb_attach(), free resources and return.
 If error occured in pq3etsec_attach(), free resources and return.
 If error occured in the attach function, free resources and return.
- If if_initialize() failed in athn_attach(), free resources and return.
- Add missing pmf_event_deregister() in athn_detach().
- Free resources correctly on some errors in atw_attach().
- Use apint*() insread of printf() in the attach function.
 If if_initialize() failed in the attach function, return.
- If if_initialize() failed in the attach function, free resources and return.
- Add missing dwc_gmac_free_dma_rings() and mutex_destroy() when attach
   failed.
- If if_initialize() failed in the attach function, free resources and return.
- ifp is always not NULL in iwi_detach(). Check correctly with ifp->if_softc.
- If if_initialize() failed in the attach function, free resources and return.
- Fix error path in the attach function correctly.
 If if_initialize() failed in the attach function, free resources and return.
 If if_attach() failed in the attach function, free resources and return.
 - If if_initialize() failed in the attach function, free resources and return.
 - KNF
 - If if_attach() failed in the attach function, free resources and return.
 - KNF
 Fix compile error.
 Fix compile error.
We don't need '&mii', but just 'mii' for mii_detach().
Don't free sc_rthash twice

Pull up following revision(s) (requested by msaitoh in ticket #424):
	sys/dev/pci/if_wm.c: 1.545-1.546
  Don't allocate MSI-X interrupt on 82583. 82583 chip has a MSI-X capability in
the PCI configuration space but it doesn't support it. At least the document
doesn't say anything about MSI-X. Fixes PR#52767 reported by
Shinichi Doyashiki.
--
- 82583 supports jumbo frame. Fixes PR#52773 reported by Shinichi Doyashiki.
- Cleanup comment.

Revert "Make if_timer MP-safe if IFEF_MPSAFE"

Because it has decreased the performance of wm. And also I found that
wm_watchdog doesn't work well with if_watchdog framework at all. Sharing one
counter (if_timer) with multiple instances (hardware multi-queues) can't detect
a single (or some) stall of them because other instances reset the counter even
if the stalled one want the watchdog to fire.

Interfaces without IFEF_MPSAFE works safely with the original if_watchdog thanks
to KENREL_LOCK. OTOH, interfaces with IFEF_MPSAFE shouldn't use if_watchdog and
should implement their own watchdog timer that works with multiple instances.

 Fix a bug that 8257[56] can't receive packet reported by Bert Kiers in
PR#52717. For 82575 and 82576, the RX descriptors must be initialized after
the setting of RCTL.EN in wm_set_filter(). This bug was added in if_wm.c
rev. 1.515.

Make if_timer MP-safe if IFEF_MPSAFE

if_timer, a counter used by if_watchdog (if_slowtimo), can be modified in
if_watchdog and if_start and/or interrupt handlers of some device drivers. All
such accesses were serialized by KERNEL_LOCK. If IFEF_MPSAFE is enabled,
KERNEL_LOCK of if_start (and perhaps interrupt handlers) is omitted and if_timer
becomes racy.

Fix the race condition by protecting if_timer by a spin mutex. if_watchdog_reset
and if_watchdog_stop are introduced to ensure to take the mutex on accessing
if_timer. Interface with IFEF_MPSAFE enabled must use the functions.

In addition, if_watchdog callout is now set CALLOUT_MPSAFE if IFEF_MPSAFE. It
means that if_watchdog implemented by a driver must be MP-safe if the driver is
set IFEF_MPSAFE.

Currenlty interfaces with IFEF_MPSAFE implementing if_watchdog and accessing
if_timer in if_start and interrupt handlers are only wm(4). wm is changed to
use the functions. (Its watchdog handler (wm_watchdog) is already MP-safe.

These contracts will be written somewhere in a further commit.

Note that the spin mutex is now ifp->if_snd.ifq_lock to avoid adding another
spin mutex to each interface. For now reusing it isn't problematic (see the
comment to know why) thought if that does matter in the future, feel free to
replace it with a new spin mutex. It's easy to do.

update from HEAD

- 82583 supports jumbo frame. Fixes PR#52773 reported by Shinichi Doyashiki.
- Cleanup comment.

 Don't allocate MSI-X interrupt on 82583. 82583 chip has a MSI-X capability in
the PCI configuration space but it doesn't support it. At least the document
doesn't say anything about MSI-X. Fixes PR#52767 reported by
Shinichi Doyashiki.

XXX pullup-8.

Pull up following revision(s) (requested by msaitoh in ticket #389):
	sys/net/if_ether.h: revision 1.69
	sys/net/if_vlan.c: revision 1.108
	sys/dev/pci/if_bge.c: revision 1.313
	sys/net/if_vlanvar.h: revision 1.11
	sys/net/if_vlanvar.h: revision 1.12
	sys/net/if_ether.h: revision 1.70
	sys/net/if_vlan.c: revision 1.110
	sys/dev/pci/if_wm.c: revision 1.544
	sys/dev/pci/if_wmreg.h: revision 1.105
  Fix a bug that a vlan packet which has priority or CFI bit in the tag causes
panic.
  Revert part of if_bge.c 1.312. It's not required to mask other than VLAN ID
bits in VLAN tag.
  Revert if_wmreg.h 1.104 and if_wm.c 1.542. It's not required to mask other
than VLAN ID bits in VLAN tag.
No functional change:
  - u_int16_t -> uint16_t
  - u_short -> uint16_t
  - tag_hash_func -> vlan_tag_hash
  - 0 -> NULL because vlr_parent is a pointer.

Pull up following revision(s) (requested by msaitoh in ticket #379):
	sys/dev/pci/if_wm.c: revision 1.540
  On I219, drop TARC0 bit 28 for DMA hang workaround (from Linux).

 Revert if_wmreg.h 1.104 and if_wm.c 1.542. It's not required to mask other
than VLAN ID bits in VLAN tag.

Unify IFEF_*_MPSAFE into IFEF_MPSAFE

There are already two flags for if_output and if_start, however, it seems such
MPSAFE flags are eventually needed for all if_XXX operations. Having discrete
flags for each operation is wasteful of if_extflags bits. So let's unify
the flags into one: IFEF_MPSAFE.

Fortunately IFEF_*_MPSAFE flags have never been included in any releases, so
we can change them without breaking backward compatibility of the releases
(though the kernel version of -current should be bumped).

Note that if an interface have both MP-safe and non-MP-safe operations at a
time, we have to set the IFEF_MPSAFE flag and let callees of non-MP-safe
opeartions take the kernel lock.

Proposed on tech-kern@ and tech-net@

Pull up following revision(s) (requested by knakahara in ticket #348):
	sys/dev/pci/if_wm.c: revision 1.542
	sys/dev/pci/if_wmreg.h: revision 1.104
fix wm(4) vlan panic. Reported and tested by Tom Ivar Helbekkmo, thanks.
wm(4) used PRI bits and CFI bit as vlantag by mistake. It is found out
by if_ether.h:r1.67.

Pull up following revision(s) (requested by kardel in ticket #1385):
	sys/dev/pci/if_wm.c: revision 1.497
PR kern/52039: use same safeguard as for the 82578

Pull up following revision(s) (requested by knakahara in ticket #302):
	sys/arch/powerpc/booke/dev/pq3etsec.c: 1.30-1.31
	sys/arch/x86/pci/if_vmx.c: 1.20
	sys/dev/ic/i82557.c: 1.148
	sys/dev/ic/rtl8169.c: 1.152
	sys/dev/pci/cxgb/cxgb_sge.c: 1.5
	sys/dev/pci/if_age.c: 1.51
	sys/dev/pci/if_alc.c: 1.25
	sys/dev/pci/if_ale.c: 1.23
	sys/dev/pci/if_bge.c: 1.311
	sys/dev/pci/if_bge.c: 1.312
	sys/dev/pci/if_bnx.c: 1.62
	sys/dev/pci/if_jme.c: 1.32
	sys/dev/pci/if_nfe.c: 1.64
	sys/dev/pci/if_sip.c: 1.167
	sys/dev/pci/if_stge.c: 1.63-1.64
	sys/dev/pci/if_ti.c: 1.102
	sys/dev/pci/if_txp.c: 1.48
	sys/dev/pci/if_vge.c: 1.61
	sys/dev/pci/if_wm.c: 1.538
	sys/dev/pci/ixgbe/ix_txrx.c: 1.29 via patch
	sys/net/agr/if_agrether_hash.c: 1.4
	sys/net/if_ether.h: 1.67-1.68
	sys/net/if_ethersubr.c: 1.244
	sys/net/if_vlan.c: 1.100
	sys/net80211/ieee80211_input.c: 1.89
	sys/net80211/ieee80211_output.c: 1.59
	sys/sys/mbuf.h: 1.171
VLAN ID uses pkthdr instead of mtag now. Contributed by s-yamaguchi@IIJ.
I just commit by proxy. Reviewed by joerg@n.o and christos@n.o, thanks.
See http://mail-index.netbsd.org/tech-net/2017/09/26/msg006459.html
--
only get vtag when we have vtag like the other drivers.
--
- only get the vtag if we have it like the other drivers
- mask the hardware vlan tag
--
- add a constant for the vlan mask.
- enforce that we have a tag before we get it.
only get vtag when we have vtag like the other drivers.
like if_bge.c:1.312 and if_stge.c:1.64.
fixed by s-yamaguchi@IIJ, thanks.

fix wm(4) vlan panic. Reported and tested by Tom Ivar Helbekkmo, thanks.

wm(4) used PRI bits and CFI bit as vlantag by mistake. It is found out
by if_ether.h:r1.67.

XXX need pullup-8

 If if_initialize() failed in the attach function, return.

 On I219, drop TARC0 bit 28 for DMA hang workaround (from Linux).

Pull up following revision(s) (requested by msaitoh in ticket #306):
	sys/dev/pci/if_wm.c: revision 1.539
- Correct NVM word size check in wm_nvm_version() to prevent
   "EERD polling failed" error on attach. if_wm.c rev. 1.534 was not perfect.
   Reported by ozaki-r@.
- Print NVM offset and word count when EERD polling failed.

- Correct NVM word size check in wm_nvm_version() to prevent
  "EERD polling failed" error on attach. if_wm.c rev. 1.534 was not perfect.
  Reported by ozaki-r@.
- Print NVM offset and word count when EERD polling failed.

XXX pullup-8.

VLAN ID uses pkthdr instead of mtag now. Contributed by s-yamaguchi@IIJ.

I just commit by proxy. Reviewed by joerg@n.o and christos@n.o, thanks.
See http://mail-index.netbsd.org/tech-net/2017/09/26/msg006459.html

XXX need pullup to -8 branch

Sync with HEAD

Pull up following revision(s) (requested by msaitoh in ticket #173):
	sys/dev/pci/if_wmreg.h: 1.99-1.103
	sys/dev/pci/if_wmvar.h: 1.34-1.37
	sys/dev/pci/if_wm.c: 1.510-1.537
	sys/dev/mii/ihphyreg.h: 1.2
	sys/dev/mii/inbmphyreg.h: 1.10
	sys/dev/mii/igphy.c: 1.27
Sync wm(4) up to if_wm.c rev. 1.537:
- Set WMREG_KABGTXD not in wm_init_locked() but in wm_reset().
- If a legacy interrupt is a spurious interrupt, don't print debug
  message.
- Don't print the Image Unique ID if an NVM is iNVM (i210 and I211).
- Fix a bug that a RAL was written at incorrect address when the index
  number is more than 16 on 82544 and newer.
- The layout of RAL on PCH* are different from others. Fix it.
- Flush every MTA write. Same as Linux.
- Move the location of calling wm_set_filter. Same as some other OSes.
- Flush writing WMREG_CTRL in wm_gmii_mediachange().
- Make new wm_phy_post_reset() and use this function at all location
  after resetting phy.
- Add I219 specific workaround for legacy interrupt. From OpenBSD.
- Move the location of calling wm_lplu_d0_disable().
- Disable D0 LPLU on 8257[12356], 82580, I350 and I21[01], too. Before
  this commit, above devices and non-PCIe devices accessed wrong
  register.
- Fix latency calculation in wm_platform_pm_pch_lpt().
- Set OBFF water mark and enable OBFF on PCH_LPT and newer.
- Fix a bug that wm_gmii_i82544_{read,write}reg() didn't take care of
  page select. PHY access from igphy() automatically did it, but
  accessing from wm(4) for workaround didn't work correctly. This
  change affects 8254[17], 8257[12] ICH8, ICH9 and ICH10.
- Call wm_kmrn_lock_loss_workaround_ich8lan() before any PHY access in
  wm_linkintr_gmii().
- Register access in wm_kmrn_lock_loss_workaround_ich8lan() now works
  correctly. Enable this function.
- IF TX is not required to flush, RX is also not required to flush in
  wm_flush_desc_rings(). Same as other OSes.
- Remove wrong semaphore access in wm_nvm_{read,write}_{ich8,spt} to
  prevent hangup. A semaphore is get/put in wm_nvm_{read,write}.
- Move some initialization stuff in wm_attach() before wm_reset().
  Some flags and callback function is required to set correctly before
  wm_reset() because wm_reset() and some helper functions refer them.
- Add wm_write_smbus_addr() to set SMBus address by software.
- Modify wm_gmii_hv_{read,write}reg_locked() to make them access
  HV_SMB_ADDR correctly.
- Configure the LCD with the extended configuration region in NVM if
  it's required. Tested with Thinkpad X220.
- 8257[12]: Don't directly access SPI but use EERD register.
- 82575-I354: If the size of SPI ROM >= 32K words, use direct SPI
  access instead of EERD register access.
- Add wm_nvm_eec_clock_raise() and wm_nvm_eec_clock_lower() and use
  them for Microwire/SPI bus control. Same as Linux and FreeBSD.
- Reduce timeout value for 80003 in wm_get_swfw_semaphore(). Same as
  Linux and FreeBSD.
- Change API of kmrn_{read,write}reg() and check the return value.
- Check whether it's required to use MDIC workaround for 80003 or not
  in wm_reset(). If the workaround isn't required, don't use the
  workaround code in wm_gmii_i80003_{read,write}reg.
- Add WM_F_WA_I210_CLSEM flag for a workaround. FreeBSD/Linux drivers
  say "In rare circumstances, the SW semaphore may already be held
  unintentionally." on I21[01]. PXE boot is one of the case.
- Qemu's e1000e emulation (82574L)'s SPI has only 64 words. I've never
  seen on real 82574 hardware with such small SPI ROM. Check
  NVM word size before accessing higher address words to prevent
  timeout.
- Check some wm_nvm_read()'s return vale.
- Use device_printf() instead of aprint_error_dev() for PHY read/write
  functions because those are used not only in device attach.
- Print internal driver flags when attaching.
- Add debug printf()s.
- Rename variables.
- Add comment, update comment and remove wrong comment.

rename unclear functions. pointed out by msaitoh@n.o.

add comment about confusing wm_rxeof() calling.

pointed out by msaitoh@n.o.

 Check some wm_nvm_read()'s return vale.

 Qemu's e1000e emulation (82574L)'s SPI has only 64 words. I've never seen on
real 82574 hardware with such small SPI ROM. Check sc->sc_nvm_wordsize before
accessing higer address words to prevent timeout.

 Add WM_F_WA_I210_CLSEM flag for a workaround. FreeBSD/Linux drivers say
"In rare circumstances, the SW semaphore may already be held unintentionally."
on I21[01]. PXE boot is one of the case.

XXX pullup-[678].

 On 8254[17] with SPI ROM, wm_nvm_set_addrbits_size_eecd() calls wm_nvm_read(),
so it's required to set flags and functions before calling it. It was broken
again in if_wm.c rev. 1.530 (see also 1.313). Tested with Dell Power Edge 1850.