Also known as CVE-2021-20269. The kdump dmesg log files(kexec-dmesg.log,
vmcore-dmesg.txt) are generated by shell redirection, which take the
default umask value, making the files readable for group and others.
This patch chmod these files, making them only accessible to owner.
Signed-off-by: Tao Liu <ltao@redhat.com>
Acked-by: Kairui Song <kasong@redhat.com>
Previously when dumping vmcore to a remote machine through ssh,
the files are created remotely and file permissions are taken
from the default umask value, which making the files accessible to
anyone on the remote machine.
This patch fixed the security issue by setting a customized umask value
before the file creation on the remote machine.
Signed-off-by: Tao Liu <ltao@redhat.com>
Acked-by: Kairui Song <kasong@redhat.com>
As FADump does not require an explicit elfcorehdr update whenever there is CPU
hotplug event so let's stop kdump service reload for FADump when CPU hotplug
event is triggered.
A new label is added to handle CPU and memory hotplug events separately. The
updated CPU hotplug event handler make sure that kdump service should not be
reloaded when FADump is configured.
Signed-off-by: Sourabh Jain <sourabhjain@linux.ibm.com>
Reviewed-by: Pingfan Liu <piliu@redhat.com>
Acked-by: Baoquan He <bhe@redhat.com>
While fadump-howto.txt talks about what happens to network interface
name on setting up a remote dump target in FADump mode, it doesn't
explicitly specify the negative consequences of it. Make it explicit
and provide a recommendation to overcome the same.
Signed-off-by: Hari Bathini <hbathini@linux.ibm.com>
Acked-by: Kairui Song <kasong@redhat.com>
On ppc64le LPAR, secure-boot is a little different from bare metal,
Where
host secure boot: /ibm,secure-boot/os-secureboot-enforcing DT property exists
while
guest secure boot: /ibm,secure-boot >= 2
Make kexec-tools adapt to LPAR
Signed-off-by: Pingfan Liu <piliu@redhat.com>
Acked-by: Kairui Song <kasong@redhat.com>
ipcalc is needed for generating 45route-static.conf. However,
on newer Fedora, e.g. 34, dracut-network drops dependency on
dhcp-client which requires ipcalc. Make kexec-tools explicitly
depends on ipcalc.
Reported-by: Jie Li <jieli@redhat.com>
Signed-off-by: Coiby Xu <coxu@redhat.com>
Acked-by: Kairui Song <kasong@redhat.com>
The kdump kernel uses resources for ftrace because trace_buf_size, which
specifies the ring buffer size for ftrace, and trace_event, which specifies
a valid trace event, are not removed, but the kdump kernel does not require
ftrace.
trace_buf_size is ignored if the specified size is 0, so specify 1.
Signed-off-by: Hisashi Nagaoka <fj1508ic@fujitsu.com>
Acked-by: Kairui Song <kasong@redhat.com>
There is requirement to decide the recommended memory size for the current
system. And the algorithm is based on /proc/iomem, so it can align with the
algorithm used by reserve_crashkernel() in kernel.
Signed-off-by: Pingfan Liu <piliu@redhat.com>
Acked-by: Kairui Song <kasong@redhat.com>
In fadump mode, it is also useful to observe kdump message through
console. Hence enable it.
Signed-off-by: Pingfan Liu <piliu@redhat.com>
Acked-by: Kairui Song <kasong@redhat.com>
At present, there is no ipv6 example for nfs and ssh dump, let's
add an example to the kdump.conf.
Signed-off-by: Lianbo Jiang <lijiang@redhat.com>
Acked-by: Pingfan Liu <piliu@redhat.com>
Currently, kdump will fail to save vmcore when using the scp and ipv6.
The reason is that the scp requires IPv6 addresses to be enclosed in
square brackets, but ssh doesn’t require this.
Let's enclose the ipv6 address in square brackets for scp dump.
Signed-off-by: Lianbo Jiang <lijiang@redhat.com>
Acked-by: Pingfan Liu <piliu@redhat.com>
There is no need to source the file manually, dracut will always
prepare the dracut lib before calling a module-setup.sh
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Lianbo Jiang <lijiang@redhat.com>
Sourcing logger file in kdump-lib.sh will leak kdump helper to dracut,
because module-setup.sh will source kdump-lib.sh. This will make kdump's
function override dracut's ones, and lead to unexpected behaviours.
So include kdump-logger.sh individually and only source it where it really
needed. for module-setup.sh, simply use dracut's logger helper is good
enough so just source kdump-logger.sh in kdump only scripts.
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Lianbo Jiang <lijiang@redhat.com>
Simplify the code and fix mount point detection. The code logic is now
much simpler: if $1 is not a mount point, call "mount --target $1" again
to try mount it. "mount --target" cmd itself can handle all the /etc/fstab
parsing job, so drop the buggy and complex bash code.
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Pingfan Liu <piliu@redhat.com>
systemctl is-failed will not work after dracut isolated to the emergency
target, so this judgement is invalid. And the restart is basically
harmless, so just revert this commit.
This reverts commit ad6a93b00d.
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Pingfan Liu <piliu@redhat.com>
That commit is trying to workaround a kernel VFS bug. Now,
the VFS issue should have been fixed in all recent releases, so
remove this workaround.
This reverts commit 539bff4083.
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Pingfan Liu <piliu@redhat.com>
Currently, the kdump sysconfig document is missed, let's add it to
the kexec-kdump-howto.txt as a document in order to help users better
understand these options in the /etc/sysconfig/kdump.
Signed-off-by: Lianbo Jiang <lijiang@redhat.com>
Acked-by: Kairui Song <kasong@redhat.com>
The current inline comment is a bit misleading, ssh dump target don't
need to use scp as core_collector, and when using scp as core_collector,
the vmcore could be huge.
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Pingfan Liu <piliu@redhat.com>
Backports:
commit 54aec3878b3f91341e6bc735eda158cca5c54ec9
Author: Alexander Egorenkov <egorenar@linux.ibm.com>
Date: Fri Sep 18 13:55:56 2020 +0200
[PATCH] make use of 'uts_namespace.name' offset in VMCOREINFO
* Required for kernel 5.11
The offset of the field 'init_uts_ns.name' has changed since
kernel commit 9a56493f6942 ("uts: Use generic ns_common::count").
Make use of the offset 'uts_namespace.name' if available in
VMCOREINFO.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Kairui Song <kasong@redhat.com>
Both $ipaddrs and $node can hold multiple strings, so use "" to brace them.
Signed-off-by: Pingfan Liu <piliu@redhat.com>
Acked-by: Kairui Song <kasong@redhat.com>
Currently, if saving vmcore failed, the final failure information won't
be saved to the kexec-dmesg.log, because the action of saving the log
occurs before the final log is printed, it has no chance to save the
log(marked it with the '^^^' below) to the log file(kexec-dmesg.log).
For example:
[1] console log:
[ 3.589967] kdump[453]: saving vmcore-dmesg.txt to /sysroot//var/crash/127.0.0.1-2020-11-26-14:19:17/
[ 3.627261] kdump[458]: saving vmcore-dmesg.txt complete
[ 3.633923] kdump[460]: saving vmcore
[ 3.661020] kdump[465]: saving vmcore failed
^^^^^^^^^^^^^^^^^^^^
[2] kexec-dmesg.log:
Nov 26 14:19:17 kvm-06-guest25.hv2.lab.eng.bos.redhat.com kdump[453]: saving vmcore-dmesg.txt to /sysroot//var/crash/127.0.0.1-2020-11-26-14:19:17/
Nov 26 14:19:17 kvm-06-guest25.hv2.lab.eng.bos.redhat.com kdump[458]: saving vmcore-dmesg.txt complete
Nov 26 14:19:17 kvm-06-guest25.hv2.lab.eng.bos.redhat.com kdump[460]: saving vmcore
Let's improve it in order to avoid the loss of important information.
Signed-off-by: Lianbo Jiang <lijiang@redhat.com>
Acked-by: Kairui Song <kasong@redhat.com>
commit 44b073b7ec467aee0d7de381d455b8ace1199184
Author: John Ogness <john.ogness@linutronix.de>
Date: Wed Nov 25 10:10:31 2020 +0106
[PATCH 2/2] printk: use committed/finalized state values
* Required for kernel 5.10
The ringbuffer entries use 2 state values (committed and finalized)
rather than a single flag to represent being available for reading.
Copy the definitions and state lookup function directly from the
kernel source and use the new states.
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Kairui Song <kasong@redhat.com>
Backports:
commit c617ec63339222f3a44d73e36677a9acc8954ccd
Author: John Ogness <john.ogness@linutronix.de>
Date: Thu Nov 19 02:41:21 2020 +0000
[PATCH 1/2] printk: add support for lockless ringbuffer
* Required for kernel 5.10
Linux 5.10 introduces a new lockless ringbuffer. The new ringbuffer
is structured completely different to the previous iterations.
Add support for retrieving the ringbuffer from debug information
and/or using vmcoreinfo. The new ringbuffer is detected based on
the availability of the "prb" symbol.
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Kazuhito Hagio <k-hagio-ab@nec.com>
Signed-off-by: Kairui Song <kasong@redhat.com>
systemctl -q --root "$initdir" add-wants X.target X.service is the
recommanded way to add service dependency, and it covers more corner
cases.
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Pingfan Liu <piliu@redhat.com>
Currently, when generating a kdump initramfs, mkdumprd will determine
how much disk space is available, if the dump target's available space
is not greater than the total system memory, mkdumprd will print a
warning to remind that there might not be enough space to save a vmcore.
Some users are complaining that mkdumprd overestimates the needed size.
But actually, the warning covers extreme scenarios such as the slab
explodes with non-zero data or a full vmcore, etc. Therefore, need to
prevent users from having minimum disk space for crash dump.
In view of this, add some descriptions to clarify it in mkdumprd man
page.
Signed-off-by: Lianbo Jiang <lijiang@redhat.com>
Acked-by: Kairui Song <kasong@redhat.com>
If dracut-initqueue failed in kdump kernel and failure action
is set to dump_to_rootfs, there is no point try again to start the
initqueue. It will also slow down the dump process, and the initqueue
will most like still not work if first attemp failed.
So just try to start sysroot.mount, if it failed, there is no luck.
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Pingfan Liu <piliu@redhat.com>
The parameter either6 is introduced to dracut by
commit 67354eebbcd4c358b8194ba5fd1ab1cf7dbd42aa
Author: Pingfan Liu <piliu@redhat.com>
Date: Tue Apr 24 16:41:21 2018 +0800
40network: introduce ip=either6 option
But it turns out needless.
On a sensible ipv6 network environment, DHCPv6 can not work properly alone,
because DHCPv6 protocol has no info about the gateway.
An reasonalbe process of ipv6 address set up should look like
host send: Router Solicitation
router reply: Router Advertisements
"Router Advertisements" carries many info like gateway, and if it has
other-config flag set, it carries DNS info etc. As for DHCPv6 address
allocation, it will only start if "Router Advertisements" has the 'managed'
flag set, which directs the host to start a stateful address allocation
from DHCPv6 server.
For more info:
rfc4861: Neighbor Discovery for IP version 6 (IPv6)
rfc5175: IPv6 Router Advertisement Flags Option
Signed-off-by: Pingfan Liu <piliu@redhat.com>
Acked-by: Kairui Song <kasong@redhat.com>
Along with 'on' option, 'fadump=' kernel parameter also supports
'nocma' & 'off' options. Update about these missing options in the
fadump-howto.txt document.
Signed-off-by: Hari Bathini <hbathini@linux.ibm.com>
Acked-by: Pingfan Liu <piliu@redhat.com>
Dracut has switch network-legacy to network-manager by default, which makes
vlan on team easy. So it can be enabled.
Testing network topology with two VMs.
VM1
ens2-\ /----> VLAN8 (192.168.120.50)
---> team0
ens3-/ (192.168.122.10)
VM2
ens2-\ /----> VLAN8 (192.168.120.100)
---> team0
ens3-/ (192.168.122.20)
Both of ens2/ens3 in VM1/VM2 are connected to virbr0.
During test, dump target is set as root@192.168.120.100:/var/crash
then crashing in VM1
Signed-off-by: Pingfan Liu <piliu@redhat.com>
Acked-by: Lianbo Jiang <lijiang@redhat.com>
Currently get_bind_mount_source will not work on btrfs, that's because
this function relies on findmnt to detect bind mount.
For a bind mount, findmnt will return different value with "-v" option.
For example, we have /dev/sdc mounted on /mnt/source, and then bind
mount /mnt/source/sub/path to /mnt/bind:
$ findmnt /mnt/bind
TARGET SOURCE FSTYPE OPTIONS
/mnt/bind /dev/sdc[/sub/path] ext4 rw,relatime,seclabel
$ findmnt -v /mnt/bind
TARGET SOURCE FSTYPE OPTIONS
/mnt/bind /dev/sdc ext4 rw,relatime,seclabel
But findmnt also return similiar result for btrfs, on a fresh installed
Fedora 33:
$ findmnt /
TARGET SOURCE FSTYPE OPTIONS
/ /dev/sdb7[/root] btrfs rw,relatime,seclabel,ssd,space_cache,subvolid=256,subvol=/root
$ findmnt -v /
TARGET SOURCE FSTYPE OPTIONS
/ /dev/sdb7 btrfs rw,relatime,seclabel,ssd,space_cache,subvolid=256,subvol=/root
The [...] indicator will contain the subvol of btrfs as well. And if
it's bind mounted under btrfs, it will contain a mixup of btrfs subvol
and the actuall fsroot.
And also, if the bind mount source device is not mounted on /,
get_bind_mount_source will also not work.
So rewrite the get_bind_mount_source function, make it work in every
cases.
Tested with:
- Silverblue's bind mount
- Bind mount with source device mounted not under /
- Btrfs
- Bind mount and source device is Btrfs
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Pingfan Liu <piliu@redhat.com>
Remove the --real when calling findmnt.
The option is only useful in capture kernel, to avoid
`findmnt` returning the pseudo 'rootfs' for non mounted path.
example, when /kdumproot/mnt/ is not mounted:
kdump:/# findmnt --target /kdumproot/mnt
TARGET SOURCE FSTYPE OPTIONS
/ rootfs rootfs rw,size=61368k,nr_inodes=15342
kdump:/# findmnt --target /kdumproot/mnt
<return 1 and empty output>
But this function will make findmnt also return empty value for bind
mount. So remove it and add an extra if statement for second kernel.
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Pingfan Liu <piliu@redhat.com>
Most watchdogs have a parameter pretimeout, if set to non-zero, it means
before the watchdog really reset the system, it will try to panic the
kernel first, so kdump could kick in, or, just print a panic stacktrace
and then kernel should reset it self.
If we are already in kdump kernel, this is not really helpful, only
increase kernel hanging chance. And it also make thing become complex
as some watchdog triggers the kernel panic in NMI context, which
could also hang the kernel in strange ways, and fail the watchdog it
self. So just disable this parameter.
Also for hpwdt, it have another parameter kdumptimeout, which is
just designed for first kernel. The default behaviour is the watchdog
will simply stop working if timeouted, trigger a panic, and leave the
kernel to kdump. Again, if we are already in kdump this is not helpful.
So also disable that.
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Lianbo Jiang <lijiang@redhat.com>
Currently the watchdog detection code is broken already, it
get the list of active watchdog drivers, then check if they are
set in the /etc/cmdline.d/* as preload module. But after we
switched to use squash module, /etc/cmdline.d/* is not directly visible.
So just detect whether current needed driver is installed.
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Lianbo Jiang <lijiang@redhat.com>
In check_fs_modified, is_nfs_dump_target is already called, the dump
target can't be nfs. No need to check here.
Signed-off-by: Kairui Song <kasong@redhat.com>
Acked-by: Lianbo Jiang <lijiang@redhat.com>
