Skip to content

Commit

Permalink
Prep for powerpc-utils-1.3.4
Browse files Browse the repository at this point in the history 
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>
  • Loading branch information
Tyrel Datwyler committed Sep 1, 2017
1 parent c8083fb commit 2e26d29
Show file tree
Hide file tree
Showing 2 changed files with 336 additions and 1 deletion.
335 changes: 335 additions & 0 deletions Changelog
View file
Original file line number Diff line number Diff line change
@@ -1,3 +1,338 @@
powerpc-utils-1.3.4
=====================================================================

commit c8083fbc1bfe1548057478ae46543bf58bffc7d4
Author: Michael Bringmann <mwb@linux.vnet.ibm.com>
Date: Wed Aug 23 09:06:45 2017 -0500

powerpc-utils.spec: Add "Vendor: IBM Corp."

Ensure that "Vendor: IBM Corp." appears in archive for all checkouts
and derived builds.

Signed-off-by: Michael Bringmann <mwb@linux.vnet.ibm.com>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 50f11c6c6618e7d59232d67e7446d07714af453b
Author: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Date: Thu Jul 6 12:31:36 2017 -0400

lsslot: Validate DRC property groups

Recently seen on a kvm guest running on a PowerNV system, the
lsslot command segfaulted when trying to free the drc information.

# gdb -q ./lsslot
Reading symbols from ./lsslot...done.
(gdb) run -c memory
Starting program: ./lsslot -c memory

Program received signal SIGSEGV, Segmentation fault.
free_drc_info () at src/drmgr/common_ofdt.c:287
287 all_drc_lists = list->all_next;

The root cause of this is that the DRC properties in the guest were empty.
This caused the lsslot (and likely possible in drmgr since this code
is common to both commands) to improperly parse the device tree
information.

To fix this we need to add a check after gathering information for
each of the DRC properties when reading DRC property group data. If
there are no entries in any of the DRC properties we cannot continue.

Reported-by: Seeteena Thoufeek <s1seetee@linux.vnet.ibm.com>
Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit c25ae395db914f1137bb8a718277ea4b6f69136b
Author: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Date: Mon Jul 10 11:23:27 2017 +1000

ppc64_cpu: Treat CPUs as online if cpu hotplug not enabled

cpu_online() looks for the "online" attribute to decide if a cpu is
online, however if the kernel does not have CONFIG_HOTPLUG_CPU enabled
the attribute will not exist.
get_attribute() calls access() first to check if the attribute exists,
which will set errno with ENOENT if it is missing. If get_attribute()
fails with this specific error, treat the cpu as online.

Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
[tyreld: removed unnecessary braces from if block]
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 817900edd96c66d6f43f75aaefbe38bba6bab035
Author: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>
Date: Fri Aug 4 16:24:01 2017 -0700

Revert "ppc64_cpu: Treat CPUs as online if cpu hotplug not enabled"

This reverts commit 5f70f6c1af2dd61122acfdbaa6e38be062bb220a.

Reverting the commit as further inspection came to show it actually
marked a cpu as offline when hotplug was not enabled.

Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 136fab970076018bd56aa1aa0c09b8954b89b0b5
Author: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Date: Tue Jun 6 12:12:36 2017 -0400

drmgr: Correct some bad formatting in drmig_chrp_pmig.c

No functional changes, justy making the code more readable.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 1c5319c76ae8b378bfd59e8c0a08e89f83867dbf
Author: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Date: Wed Jun 7 16:35:33 2017 -0400

drmgr: Correct errno usage in init_cpu_info()

The error message reported if allocaing a dr node fails
should not report the errno value. The call to allocate the
dr node makes multiple syscalls and there is no way to
determine which one last updated errno.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Reviewed-by: Paul Clarke <pc@us.ibm.com>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 998933a5185f9c56d1d6bce77301f3e7be6a54cf
Author: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Date: Wed Jun 7 16:35:28 2017 -0400

drmgr: Correct errno usage use in validate_paltform()

For validate_platform(), the call to get_property makes many
system calls and there is no way to know what last updated the
errno value. This patch drops the output of strerror() due to
this reason.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Reviewed-by: Paul Clarke <pc@us.ibm.com>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit d1c6662d38adeb98a1fd8906cd54031c6fdd653d
Author: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Date: Wed Jun 7 16:35:22 2017 -0400

drmgr: Stale errno usage corrections

Several places in the drmgr code use a stale value of errno either
as a return code or when printing an error message. A common example
of this is calling the say() routine and then returning the value of
errno. The say() routine makes several system calls and the value of
errno after a call to say() no longer reflects the original errno
value.

Several places in the drmgr code were updated to properly handle
the return codes from functions such as write() or open() where we
should be checking for rc == -1 instead of rc <= 0. This does make
a difference as errno won't be set for rc == 0.

Note that partial write return codes are not really handled in the
drmgr code. The drmgr command is writing to sysfs and /proc
files where we know that they will either return success or
failure and set errno, no partial writes.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Reviewed-by: Paul Clarke <pc@us.ibm.com>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 479f1a1122f7d9025d3a21e60a0431d2eba6ee2f
Author: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Date: Tue Jun 6 12:12:22 2017 -0400

drmgr: Correct drc type checking in check_kmods()

Fixes: f9986ac43011 ("drmgr: Move options.ctype to global usr_drc_type")

When the conversion to using the global usr_drc_type was
made the logic checking for whic drc types we need to check
the rpadlpar_io kernbel module for was broken. This corrects
the check.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 27e8d5b8218bc8c8c772080b0d99691bd31d4f0f
Author: Chandni Verma <chandni@linux.vnet.ibm.com>
Date: Thu Jun 1 19:07:13 2017 +0530

lsslot: Rectify usr_drc_name check

This incorrect check refrains lsslot from displaying the slot information.

Fixes: e82fb588f573 (lsslot: replace struct options member s_name with usr_drc_name)
Signed-off-by: Chandni Verma <chandni@linux.vnet.ibm.com>
Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
[tyreld: added fixes tag]
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 74715cc0e4db0e2738be7a19c852816c285d8284
Author: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Date: Mon May 22 16:40:58 2017 -0400

lsdevinfo: Add support for vNIC devices

Add support to dump information about vNIC devices so that device info
requests from the HMC can retrieve data about vNIC devices.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
[tyreld: fixed up trailing whitespace]
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 25615a80f3fb5ebcc26d7b00316afce8c46ab2f9
Author: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Date: Mon May 22 16:40:50 2017 -0400

ofpathname: Add support for vNIC devices

Add 'vnic' to the list of device types for of2l_ethernet so vNIC
devices are supported.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 6464f8e96228f7db355f08f270ada1104c3562c3
Author: Amartey Pearson <apearson@us.ibm.com>
Date: Tue May 16 14:01:30 2017 -0500

Improve perf of drmgr/lsslot with large num of virt. dev.

The current algorithm for add_pci_vio_node walks the vdevice bus
for every theoretical DRC (1000's of times). The more vdevices
you have populated, the worse this algorithm becomes. On a system
with 450 devices, this can take over 30 seconds for example.

This modifies the algorithm to populate a node list of existing
devices (so walks the tree just once), and then uses these nodes
when populating the full list instead of walking the device tree
for each one.

NOTE: This only is changed for virtual devices - a system with
large numbers of PCI devices will not be optimized, but that is
less of a concern.

Signed-off-by: Amartey Pearson <apearson@us.ibm.com>
Signed-off-by: Bryant G. Ly <bryantly@linux.vnet.ibm.com>
[tyreld: changed my_drc_index type from u32 to uint32_t]
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 7539330a301ff9ac7d6fd5e1cf974bb9999100b2
Author: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>
Date: Fri May 19 16:02:51 2017 -0400

Test for zlib presence at configuration time

The nvram utility links with zlib and compilation will fail if zlib.h or
libz.so are not present. Check and warn at configuration time.

Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 3a7b15104fb956aa2e5debababe8536dbd4e8cef
Author: Russell Currey <ruscur@russell.cc>
Date: Fri May 19 17:36:45 2017 -0400

errinjct: Remove platform check to enable KVM guests

qemu/KVM guests can have RTAS as well. There's already a check to see if
RTAS is present, there's no need to hardcode the supported platform.
Removing this check allows errinjct to work on KVM guests.

Signed-off-by: Russell Currey <ruscur@russell.cc>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 5f70f6c1af2dd61122acfdbaa6e38be062bb220a
Author: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Date: Fri Apr 21 16:47:57 2017 +1000

ppc64_cpu: Treat CPUs as online if cpu hotplug not enabled

cpu_online() looks for the "online" attribute to decide if a cpu is
online, however if the kernel does not have CONFIG_HOTPLUG_CPU enabled
the attribute will not exist.
If hotplug is not enabled then the cpu can be assumed to be online - in
this case, update cpu_online() to treat the cpu as online.

Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Acked-by: Stewart Smith <stewart@linux.vnet.ibm.com>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 4fcc611a09caea70e6bca6a1275211c72ac4bb45
Author: Seeteena Thoufeek <s1seetee@linux.vnet.ibm.com>
Date: Wed Jan 11 15:36:33 2017 +0530

warning: command substitution: ignored null byte in input device

Ubuntu 17.04: "command substitution: ignored null byte in input" warning message is displayed while running few RAS commands.

The command execution of ls-veth, ls-vscsi, lsdevinfo, and ofpathname returns following output.

lsdevinfo
/usr/sbin/lsdevinfo: line 233: warning: command substitution: ignored null byte in input
device: name="ibmveth2"
ls-vscsi
/usr/sbin/ls-vscsi: line 72: warning: command substitution: ignored null byte in input
host0 U8286.42A.1069B3T-V8-C17-T1
ls-veth
/usr/sbin/ls-veth: line 74: warning: command substitution: ignored null byte in input
ibmveth2 U8286.42A.1069B3T-V8-C2-T1
ofpathname sda
/usr/sbin/ofpathname: line 812: warning: command substitution: ignored null byte in input
/usr/sbin/ofpathname: line 865: cd: too many arguments
/vdevice/v-scsi@30000011/disk@8100000000000000

Newer versions of bash (Bash 4.4) warn about null bytes in command substitutions. This is fixed.

Also fixed /usr/sbin/ofpathname: line 865: cd: too many arguments.

Signed-off-by: Seeteena Thoufeek <s1seetee@linux.vnet.ibm.com>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

commit 25b8f7d67fb3acee0970cd36f24dc17fac1f6dc3
Author: Pawan K Singh5 <pawan123@linux.vnet.ibm.com>
Date: Mon Apr 10 15:00:28 2017 +0530

ofpathname : avoid duplicate entries of vFC devices

Adding support to avoid the duplicate entries of a Open Firmware device path to
logical device path for vfc-client devices.

The problem can be reproduced with the bootlist command

# bootlist -m normal /dev/sdd /dev/sds /dev/sdf /dev/sdv
# bootlist -m normal -r
/vdevice/vfc-client@30000899/disk@500507680b2255fe
/vdevice/vfc-client@30000899/disk@500507680b2255ff
/vdevice/vfc-client@3000089a/disk@500507680b2255fe
/vdevice/vfc-client@3000089a/disk@500507680b2255ff
# bootlist -m normal -o
sda---------------->expected "sdd"
sdp---------------->expected "sds"
sda---------------->expected "sdf"
sdp---------------->expected "sdv"

So, this patch check the duplicate entries of logical names of vFC devices at of2l_vfc()
and display specified bootlist :

# bootlist -m normal -r
/vdevice/vfc-client@30000899/disk@500507680b2255fe
/vdevice/vfc-client@30000899/disk@500507680b2255ff
/vdevice/vfc-client@3000089a/disk@500507680b2255fe
/vdevice/vfc-client@3000089a/disk@500507680b2255ff
# bootlist -m normal -o
sdd
sds
sdf
sdv

Signed-off-by: Pawan K Singh5 <pawan123@linux.vnet.ibm.com>
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>

powerpc-utils-1.3.3
=====================================================================

Expand Down
2 changes: 1 addition & 1 deletion configure.ac
View file
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
# -*- Autoconf -*-
# Process this file with autoconf to produce a configure script.
m4_define([ppu_version], 1.3.3)
m4_define([ppu_version], 1.3.4)

AC_PREREQ([2.63])
AC_INIT([powerpc-utils], ppu_version, [nfont@linux.vnet.ibm.com])
Expand Down

0 comments on commit 2e26d29

Please sign in to comment.