[SRU][J][F][PATCH 0/8] Problems with HVCS and hotplugging (LP: 2056373)

frank.heimes at canonical.com frank.heimes at canonical.com
Fri Mar 8 12:28:12 UTC 2024


BugLink: https://bugs.launchpad.net/bugs/2056373

SRU Justification:

[Impact]

 * HVCS (Hypervisor Virtual Console Server) is broken because the
   virtual terminal mkvterm fails, caused by pvmutil failing.

 * When mkvterm is ran, it ultimately fails because it calls pvmutil
   which fails.
   pvmutil calls drmgr, and drmgr is adding a slot correctly.
   However, when drmgr writes the slot information to ?/add_slot,
   the return is -ENODEV.

 * This leads to HVCS never having probe() called.

 * In addition, HVCS is missing patches/fixes, and is broken without them.

[Fix]

 * Fix one and two is required for focal only, all other for focal and jammy:

 * 57409d4fb12c 57409d4fb12c185b2c0689e0496878c8f6bb5b58
   "powerpc/pseries: Fix bad drc_index_start value parsing of drc-info entry"

 * c5e76fa05b2d c5e76fa05b2df519b9f08571cc57e623c1569faa
   "powerpc/pseries: Fix of_read_drc_info_cell() to point at next record"

 * 6a9a733edd46 6a9a733edd46732e906d976dc21a42dd361e53cc
   "hvcs: Fix hvcs port reference counting"

 * 760aa5e81f33 760aa5e81f33e0da82512c4288489739a6d1c556
   "hvcs: Use dev_groups to manage hvcs device attributes"

 * 503a90dd619d 503a90dd619d52dcac2cc68bd742aa914c7cd47a
   "hvcs: Use driver groups to manage driver attributes"

 * 3a8d3b366ce4 3a8d3b366ce47024bf274eac783f8af5df2780f5
   "hvcs: Get reference to tty in remove"

 * d432228bc7b1 d432228bc7b1b3f0ed06510278ff5a77b3749fe6
   "hvcs: Use vhangup in hotplug remove"

 * 28d49f8cbe9c 28d49f8cbe9c7966f91ee1b5ec2f997f6e55bf9f
   "hvcs: Synchronize hotplug remove with port free"

[Test Plan]

 * The high level test plan is to run mkvterm with an id.
 
 * mkvterm will fail because /dev/hvcs* device nodes are missing.

 * Details see https://bugs.launchpad.net/bugs/2023243 for more information.
   Especially the script provided by IBM
   (see original bug description: `---Steps to Reproduce---`).

 * IBM will (stress) test the updated kernel(s) provided in -proposed.

[Where problems could occur]

 * The first two commits affect arch/powerpc/platforms/pseries/of_helpers.c
   and are needed to fix the hotplugging issue seen when drmgr goes to write
   the slot information to /sys/bus/pci/slots/control/add_slot.
   In case of issues here hotplugging with drmgr might break.

 * The issue lies in rpadlpar_io and rpaphp calling an of helper function
   of_read_drc_info_cell(). Without these commits, the value stored
   drc_index_start is incorrect.
   This ultimately results in the entire SLOT string being incorrect,
   and rpaphp never finding the newly added slot by drmgr.
   rpadlpar then returns -ENODEV.
   Therefore, HVCS is never probed, and the device nodes are never created.

 * HVCS, rpadlpar_io, and rpaphp should ideally not even need to be loaded
   prior to drmgr adding a vio slot.
   If rpadlpar_io and rpaphp are not loaded, drmgr will load them.
   In addition, if rpadlpar_io and rpaphp register the new slot correctly,
   rpadlpar_io will call dlpar_add_vio_slot(),
   which calls vio_register_device_node() with the device node.
   This is what tells the driver core to init and probe HVCS
   (which is needed to create the device nodes).

 * The remaning 6 commits are needed for HVCS, that is essentially
   broken without them.
   Overall, issues they fix are race conditions, hotplug remove issues,
   as well as memory leaks.

 * Please notice that this is entirely ppc64el architecture-specifc.

[Other Info]

 * All the commits listed above are included in mantic and noble.
   Hence these are set to Invalid.

 * Meanwhile these requested commits have been added to other
   kernels and distros.

Brian King (6):
  hvcs: Fix hvcs port reference counting
  hvcs: Use dev_groups to manage hvcs device attributes
  hvcs: Use driver groups to manage driver attributes
  hvcs: Get reference to tty in remove
  hvcs: Use vhangup in hotplug remove
  hvcs: Synchronize hotplug remove with port free

Tyrel Datwyler (2):
  powerpc/pseries: Fix bad drc_index_start value parsing of drc-info
    entry
  powerpc/pseries: Fix of_read_drc_info_cell() to point at next record

 arch/powerpc/platforms/pseries/of_helpers.c | 10 +--
 drivers/tty/hvc/hvcs.c                      | 91 ++++++++-------------
 2 files changed, 38 insertions(+), 63 deletions(-)

-- 
2.25.1




More information about the kernel-team mailing list