kernel-packages team mailing list archive

[bugproxy <bugproxy@xxxxxxxxxx>]

------- Comment From hbathini@xxxxxxxxxx 2016-04-12 07:47 EDT-------
(In reply to comment #8)
> Hello,
>

Hi Louis,
> I am not against the modification but would like to understand better the
> rationale behind this. According to the kernel doc we have :

The rationale behind moving from maxcpus=1 to nr_cpus=1 is to reduce
the memory consumption in kdump kernel. nr_cpus is a hard limit that
has an impact on the (kdump) kernel memory consumption,
while it is not the case with maxcpus=1, as we can theoretically hotplug cpus
with maxcpus=1  (not something we would be needing for kdump).

> maxcpus=        [SMP] Maximum number of processors that an SMP kernel
> should make use of.  maxcpus=n : n >= 0 limits the
> kernel to using 'n' processors.  n=0 is a special case,
> it is equivalent to "nosmp", which also disables
> the IO APIC.
>
> nr_cpus=        [SMP] Maximum number of processors that an SMP kernel
> could support.  nr_cpus=n : n >= 1 limits the kernel to
> supporting 'n' processors. Later in runtime you can not
> use hotplug cpu feature to put more cpu back to online.
> just like you compile the kernel NR_CPUS=n
>
> I personally would be inclined to use maxcpus=0 and totally disable SMP but
> I am opened to other options.
>

As for maxcpus=0, I am not really sure if powerpc supports maxcpus=0 / nosmp
currently.

Thanks
Hari

-- 
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to makedumpfile in Ubuntu.
https://bugs.launchpad.net/bugs/1568952

Title:
  ISST-LTE:pVM:thymelp2:ubuntu 16.04: change "maxcpus=1" to "nr_cpus=1"
  in kdump-tools

Status in makedumpfile package in Ubuntu:
  New

Bug description:
  == Comment: #0 - Ping Tian Han - 2016-04-07 23:04:30 ==
  ---Problem Description---
  Because canonical  has fixed bug 137281 ( LP: #1560552 ) in 4.4.0-17-generic, the kernel now supports "nr_cpus=1". So we should update the "maxcpus=1" to "nr_cpus=1" in /etc/default/kdump-tools I think.
   
  ---uname output---
  Linux thymelp2 4.4.0-17-generic #33-Ubuntu SMP Tue Mar 29 17:15:31 UTC 2016 ppc64le ppc64le ppc64le GNU/Linux
   
  Userspace tool: kdump-tools  version 1:1.5.9-5 

  == Comment: #2 - Hari Krishna Bathini - 2016-04-11 04:36:14 ==
  (In reply to comment #1)
  > Hari.
  > 
  > Would you please validate that nr_cpus is consistent across architectures
  > and should become the new default?
  > 
  > Thanks.

  Hi Kevin,

  I know nr_cpus=1 is supported on x86 based on this patch
  https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=14cb6dcf0a023f5977461c94d8d5a163c937979b

  About other architectures, I am not really sure.

  But looking at the kdump-tools source package, it seems like other architectures
  are using maxcpus=1 to boot kdump kernel unless there is architecture specific
  overrides that are missing in the source. Canonical might be able to answer better.

  If other aren't using or willing to use nr_cpus=1, we may need a patch
  of this kind:

  diff --git a/kdump-config b/kdump-config
  index 0ff0e6f..aba300e 100755
  --- a/kdump-config
  +++ b/kdump-config
  @@ -48,7 +48,11 @@ KDUMP_COREDIR=${KDUMP_COREDIR:=/var/crash}
   KDUMP_DUMP_DMESG=${KDUMP_DUMP_DMESG:=1}
   KDUMP_DIR="/var/lib/kdump"
   MAKEDUMP_ARGS=${MAKEDUMP_ARGS:="-c -d 31"}
  -KDUMP_CMDLINE_APPEND=${KDUMP_CMDLINE_APPEND:="irqpoll maxcpus=1 nousb systemd.unit=kdump-tools.service"}
  +if [ "$ARCH" = "ppc64le" ]; then
  +       KDUMP_CMDLINE_APPEND=${KDUMP_CMDLINE_APPEND:="irqpoll nr_cpus=1 nousb systemd.unit=kdump-tools.service"}
  +else
  +       KDUMP_CMDLINE_APPEND=${KDUMP_CMDLINE_APPEND:="irqpoll maxcpus=1 nousb systemd.unit=kdump-tools.service"}
  +fi
   KDUMP_KERNEL_HOOK="/etc/kernel/postinst.d/kdump-tools"
   [ -d $KDUMP_COREDIR ] || mkdir -p $KDUMP_COREDIR ;

  == Comment: #4 - Kevin W. Rudd - 2016-04-11 11:08:16 ==
  Mirroring to Canonical for their review and feedback.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/makedumpfile/+bug/1568952/+subscriptions