Hi,
About this bug, I think the root cause is the definition of option "-p"
is not very clear. In man page, "-p" means "Load the new kernel for use
on panic." But when do the crsh unloading, "-p" also need be specified
with "-p". This is maing the "-p" confusing. Why an unloading job needs
a loading option.
For this issue, maybe 3 choices can be taken:
1, like this patch has done, it works and fix the error.
2, Just change the description in man page and help message, note people
"-p -u" have to be specified and keep the strict sequence. Exchange will
cause incorrect consequence.
3, Fix the root cause. Make clear the meaning of "-p", it just means the
attribute of action. If with "-p", an crash related job will be done, as for
the action, you need specify "-l" or "-u". If only "-l" and "-u", it's
the loading/unloading actions for kexec. If "-p" is added with "-l" and
"-u", then it's the loading/unloading actions for panic. Then in code
parsing the "-p" option, it can be changed like below:
diff --git a/kexec/kexec.c b/kexec/kexec.c
index 7e7b604..9b5aa96 100644
--- a/kexec/kexec.c
+++ b/kexec/kexec.c
@@ -1277,9 +1277,6 @@ int main(int argc, char *argv[])
type = optarg;
break;
case OPT_PANIC:
- do_load = 1;
- do_exec = 0;
- do_shutdown = 0;
do_sync = 0;
if (do_kexec_file_syscall)
kexec_file_flags |= KEXEC_FILE_ON_CRASH;
With this change, if user want to do a panic loading/unloading, he/she
need be specify
"kexec -p -l /boot/vmlinuz-xxx initrd /boot/initramfs-xx ....."
"kexec -u -p"
So what's you guys' opinion?
Thanks
Baoquan
On 08/19/14 at 04:45pm, Baoquan He wrote:
> Currently, unload work will be done successfully when specify "kexec -p -u".
> However, when exchange the sequence like "kexec -u -p" unload work can be
> done too, but with a usage printing.
>
> This is because in the former sequence, "-p" will do below assignments:
> do_load=0;
> do_unload=1;
> Then "-u" do below assignments.
> do_load=0;
> do_unload=1;
> Nothing wrong happened in this situation. Since the result woule be:
> "do_load==0" and do_unload==1.
>
> In the latter sequence, the final result will be "do_load==1" and
> "do_unload==1". Then after unload operation, it will go into loading
> code flow. Thsi cause the usage() calling and printing.
>
> So in this patch add a check when enter into the loading code flow.
>
> Signed-off-by: Baoquan He <bhe at redhat.com>
> ---
> kexec/kexec.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kexec/kexec.c b/kexec/kexec.c
> index 7e7b604..87ec26a 100644
> --- a/kexec/kexec.c
> +++ b/kexec/kexec.c
> @@ -1361,7 +1361,7 @@ int main(int argc, char *argv[])
> else
> result = k_unload(kexec_flags);
> }
> - if (do_load && (result == 0)) {
> + if (do_load && !do_unload (result == 0)) {
> if (do_kexec_file_syscall)
> result = do_kexec_file_load(fileind, argc, argv,
> kexec_file_flags);
> --
> 1.8.5.3
>
