[Ubuntu 18.04.1] POWER9 - Nvidia Volta - Kernel changes to enable Nvidia driver on bare metal

[bugproxy]

== SRU Justification ==
12 kernel patches have been identified as needed to support Nvidia Volta
on bare metal. All are accepted upstream in 4.17. Three of those are
already in bionic, leaving a total of 9 remaining commits needed in bionic.
This pull request is for those other 9 commits.

== Regression Potential ==
All of the commits are specific to powerpc.

== Test Case ==
A test kernel was built with these patches and tested by IBM.

== Comment: #0 - Barry B. Arndt <email address hidden> - 2018-05-23 13:40:33 ==
12 kernel patches have been identified as needed to support Nvidia Volta on bare metal. All are accepted upstream in 4.17. Three of those are already in bionic, leaving a total of 9 remaining commits needed in bionic. Those 9 commits are:

720c84046c26 powerpc/npu-dma.c: Fix crash after __mmu_notifier_register failure
2b74e2a9b39d powerpc/powernv/npu: Fix deadlock in mmio_invalidate()
5ee573e8ef03 powerpc/powernv/mce: Don't silently restart the machine
fb5924fddf9e powerpc/mm: Flush cache on memory hot(un)plug
7fd6641de28f powerpc/powernv/memtrace: Let the arch hotunplug code flush cache
28a5933e8d36 powerpc/powernv/npu: Add lock to prevent race in concurrent context init/destroy
a1409adac748 powerpc/powernv/npu: Prevent overwriting of pnv_npu2_init_contex() callback parameters
d0cf9b561ca9 powerpc/powernv/npu: Do a PID GPU TLB flush when invalidating a large address range
75ecfb49516c powerpc/mce: Fix a bug where mce loops on memory UE.

We cherry-picked the commits, and all applied cleanly. The resultant kernel built successfully and loaded.

[Joseph Salisbury (jsalisbury)]

I built a Bionic test kernel with the 9 commits posted in the bug description. The test kernel can be downloaded from:
http://kernel.ubuntu.com/~jsalisbury/lp1772991

Can you test this kernel and see if it resolves this bug?

Note about installing test kernels:
• If the test kernel is prior to 4.15(Bionic) you need to install the linux-image and linux-image-extra .deb packages.
• If the test kernel is 4.15(Bionic) or newer, you need to install the linux-modules, linux-modules-extra and linux-image-unsigned .deb packages.

Thanks in advance!

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2018-05-24 10:20 EDT-------
HI Canonical,

Fred (thank!) tested this kernel and it is working fine. He was even able to install CUDA stack on AC922 and it worked great!

[Joseph Salisbury (jsalisbury)]

Thanks for testing. SRU request submitted:
https://lists.ubuntu.com/archives/kernel-team/2018-May/092729.html

[Breno Leitão (breno-leitao)]

hi Joseph,

I tried to discover if the patchset above was accepted/acked but I didn't find anything. Do you know if the patchset would make the SRU criteria?

[Joseph Salisbury (jsalisbury)]

We are still waiting for a couple of acks at this point. The patch set should meeting the SRU criteria, just waiting for review. I'll resend the SRU request if the acks don't happen soon.

[Joseph Salisbury (jsalisbury)]

These patches are now applied to the Bionic master-next repo.

[Brad Figg (brad-figg)]

This bug is awaiting verification that the kernel in -proposed solves the problem. Please test the kernel and update this bug with the results. If the problem is solved, change the tag 'verification-needed-bionic' to 'verification-done-bionic'. If the problem still exists, change the tag 'verification-needed-bionic' to 'verification-failed-bionic'.

If verification is not done by 5 working days from today, this fix will be dropped from the source code, and this bug will be closed.

See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you!

[Kleber Sacilotto de Souza (kleber-souza)]

Hello IBM,

Could you please verify the fix(es) with the Bionic kernel currently in -proposed?

Thank you.

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2018-06-25 09:59 EDT-------
I am working on this and plan to finish today or tomorrow (6/26).

[Douglas Lehr (dllehr-ibm)]

I can verify from IBM's standpoint, that we were able to use Volta gpu's with POWER9. The Nvidia driver enabled correctly.

[Kleber Sacilotto de Souza (kleber-souza)]

Thank you Douglas for the verification!

[Launchpad Janitor (janitor)]

This bug was fixed in the package linux - 4.15.0-24.26

---------------
linux (4.15.0-24.26) bionic; urgency=medium

  * linux: 4.15.0-24.26 -proposed tracker (LP: #1776338)

  * Bionic update: upstream stable patchset 2018-06-06 (LP: #1775483)
    - drm: bridge: dw-hdmi: Fix overflow workaround for Amlogic Meson GX SoCs
    - i40e: Fix attach VF to VM issue
    - tpm: cmd_ready command can be issued only after granting locality
    - tpm: tpm-interface: fix tpm_transmit/_cmd kdoc
    - tpm: add retry logic
    - Revert "ath10k: send (re)assoc peer command when NSS changed"
    - bonding: do not set slave_dev npinfo before slave_enable_netpoll in
      bond_enslave
    - ipv6: add RTA_TABLE and RTA_PREFSRC to rtm_ipv6_policy
    - ipv6: sr: fix NULL pointer dereference in seg6_do_srh_encap()- v4 pkts
    - KEYS: DNS: limit the length of option strings
    - l2tp: check sockaddr length in pppol2tp_connect()
    - net: validate attribute sizes in neigh_dump_table()
    - llc: delete timers synchronously in llc_sk_free()
    - tcp: don't read out-of-bounds opsize
    - net: af_packet: fix race in PACKET_{R|T}X_RING
    - tcp: md5: reject TCP_MD5SIG or TCP_MD5SIG_EXT on established sockets
    - net: fix deadlock while clearing neighbor proxy table
    - team: avoid adding twice the same option to the event list
    - net/smc: fix shutdown in state SMC_LISTEN
    - team: fix netconsole setup over team
    - packet: fix bitfield update race
    - tipc: add policy for TIPC_NLA_NET_ADDR
    - pppoe: check sockaddr length in pppoe_connect()
    - vlan: Fix reading memory beyond skb->tail in skb_vlan_tagged_multi
    - amd-xgbe: Add pre/post auto-negotiation phy hooks
    - sctp: do not check port in sctp_inet6_cmp_addr
    - amd-xgbe: Improve KR auto-negotiation and training
    - strparser: Do not call mod_delayed_work with a timeout of LONG_MAX
    - amd-xgbe: Only use the SFP supported transceiver signals
    - strparser: Fix incorrect strp->need_bytes value.
    - net: sched: ife: signal not finding metaid
    - tcp: clear tp->packets_out when purging write queue
    - net: sched: ife: handle malformed tlv length
    - net: sched: ife: check on metadata length
    - llc: hold llc_sap before release_sock()
    - llc: fix NULL pointer deref for SOCK_ZAPPED
    - net: ethernet: ti: cpsw: fix tx vlan priority mapping
    - virtio_net: split out ctrl buffer
    - virtio_net: fix adding vids on big-endian
    - KVM: s390: force bp isolation for VSIE
    - s390: correct module section names for expoline code revert
    - microblaze: Setup dependencies for ASM optimized lib functions
    - commoncap: Handle memory allocation failure.
    - scsi: mptsas: Disable WRITE SAME
    - cdrom: information leak in cdrom_ioctl_media_changed()
    - m68k/mac: Don't remap SWIM MMIO region
    - block/swim: Check drive type
    - block/swim: Don't log an error message for an invalid ioctl
    - block/swim: Remove extra put_disk() call from error path
    - block/swim: Rename macros to avoid inconsistent inverted logic
    - block/swim: Select appropriate drive on device open
    - block/swim: Fix array bounds check
    - block/swim: Fix IO error at end of medium
    -...