On 03/24/2017 at 04:27 PM, Dave Young wrote:
On 03/23/17 at 04:47pm, Xunlei Pang wrote:
> We add a generic dump target hook for kdump, supposed to monitor all the
> devices mounted by kdump.
> PATCH6 implements the basic framework, add a hook kdump_target_hook(),
> and provide a generic helper kdump_target_check() for use(must use
> it under mkdump/dracut).
> The first five patches are prerequisites for the mechanism, this will
> ensure us only have one device(our dump target) mounted after kdump boot,
> there still are corner cases like the special "x-initrd.mount" in
> we can discuss it further to see if we should mount it by kdump or handle it?
> PATCH7 is an application of this mechanism(add the "lvm_linear"
> if the dump target is linear type lvm volume, we reduce the
> in "/etc/lvm/lvm.conf" from 8MB to 1MB(we discussed this with lvm team,
> agree that we can safely set it to 1MB in case of only one lvm linear target).
> PATCH8 is another application(add the "lvm" checklist) to remove
> module when no lvm target needed by kdump.
> Xunlei Pang (8):
> kdumpctl: fix a bug in remove_cmdline_param()
> kdumpctl: remove "root=X" for kdump boot
> kdumpctl: fix the rebuild issue after removing "root=X"
> Revert "kdumpctl: filter 'root' kernel parameter when running in live
> kdumpctl: move is_fadump_capable() into kdump-lib.sh
> mkdumprd: add a generic hook for the dump target
> mkdumprd: reduce lvm2 memory in case of linear volume
> mkdumprd: omit "lvm" dracut module if no lvm target
> dracut-module-setup.sh | 6 +++
> dracut-process-lvmconf.sh | 6 +++
> kdump-lib.sh | 25 +++++++++
> kdumpctl | 36 +++++--------
> kexec-tools.spec | 3 ++
> mkdumprd | 128 +++++++++++++++++++++++++++++++++++++++-------
> 6 files changed, 163 insertions(+), 41 deletions(-)
> create mode 100644 dracut-process-lvmconf.sh
> kexec mailing list -- kexec(a)lists.fedoraproject.org
> To unsubscribe send an email to kexec-leave(a)lists.fedoraproject.org
Have not finished read all the patches, but seem this series is doing
two things, one is remove root dependencies, another is shrink lvm
memory usage. How about split them into two series. We can review them
separately considering the lvm part is a little complicated.
Thanks for your comment.
Yes, they are two things. But "lvm" patches rely on "remove root
ensures that we only have one target, but it is a little hard. As I replied the
it's actually related(host_devs doesn't need to be mounted, but still may be a
In other words, we have to make sure that all the host devices added to kdump are hooked.
For example, in case of nfs dumping, the current implementation still has root, it the
lvm volume as it normally is, we will have a wrong judgment. For this point, I guess I can
to add an extra hook for the root device to split it into two series.
BTW, seems patch 2 and patch 3 can be merged in one patch.
Yes, can be merged.