Bug 1132300

Description of problem:
I know you folks don't support custom kernels, and this corner case will probably won't present itself in enterprise environment, however since the
bug is in user space script, and custom compiled kernel only created a 
condition that's not handled properly by mkdumprd script I feel obligated to report this issue and let you folks decide.

Given that "/proc/sys/crypto/fips_enabled" is not present (Some crypto library is probably not enabled).

Issue 1:
When you start the kdump service when no kdump initial ramdisk is present, it tries to rebuild the ramdisk and calls script mkdumprd at some point.

[root@localhost ~]# service kdump start
No kdump initial ramdisk found.                            [WARNING]
Rebuilding /boot/initrd-2.6.34-rckdump.img
cat: /proc/sys/crypto/fips_enabled: No such file or directory
/sbin/mkdumprd: line 3923: [: : integer expression expected
Starting kdump:                                            [  OK  ]

[root@localhost ~]# echo $?
0

In the absence of fips_enabled file, variable FIPS_MODE is not set.
While the comparison [ "$FIPS_MODE" -eq 1 ] && setup_fips on line 3923
would have worked fine for string comparison regardless of whether variable is set/unset, it doesn't for integer comparison when variable is not set. Script moves ahead regardless, and creates initial ramdisk.

I copied the script, and wrote wrote two patches that takes care of "/sbin/mkdumprd: line 3923: [: : integer expression expected" warning, but as far as "cat: /proc/sys/crypto/fips_enabled: No such file or directory" message is concerned, I don't know how you folks want to handle that.

Patch 1:
--- /sbin/mkdumprd      2014-08-21 10:37:30.848654720 +0530
+++ /root/mkdumprd      2014-08-21 10:33:07.819945059 +0530
@@ -3920,7 +3920,7 @@
     done | sort | uniq | sed '/^ *$/d'`
     rm -f $TEMPLDCFG

-    [ "$FIPS_MODE" -eq 1 ] && setup_fips
+    [ "0$FIPS_MODE" -eq 1 ] && setup_fips   // it's numeric zero not the                           alphabet O

     #copy the binaries and their shared libraries to the archive
     for n in $bin $kdump_libs $k_extras $fips_hmac; do

Patch 2:
--- /sbin/mkdumprd      2014-08-21 10:43:31.793620929 +0530
+++ /root/mkdumprd      2014-08-21 10:43:25.312605379 +0530
@@ -93,7 +93,7 @@
 DISK_TIMEOUT=180
 DEBUG_MEM_LEVEL="0"

-FIPS_MODE=$(cat /proc/sys/crypto/fips_enabled)
+FIPS_MODE=$(cat /proc/sys/crypto/fips_enabled || echo -n 0)

 error()
 {

Issue 2:
Run /sbin/mkinitrd --help
cat: /proc/sys/crypto/fips_enabled: No such file or directory
usage: mkdumprd [--version] [--help] [-v] [-d] [-f] [--preload <module>]
       [--image-version]
       [--builtin=<module>] [--omit-dmraid]
       [--fstab=<fstab>] [--nocompress] <initrd-image> <kernel-version>

       (ex: mkdumprd /boot/initrd-2.2.5-15.img 2.2.5-15)

It should not be checking for fips_enabled file in first place if invoked with --help option.
 
Version-Release number of selected component (if applicable):
kexec-tools-2.0.0-273.el6.x86_64


How reproducible: Since you can't delete "fips_enabled" file from proc, and compiling you own kernel without fips may be too much work, you can easily reproduce the bug if you 
1. Change the path "/proc/sys/crypto/fips_enabled" in line 96 of file /sbin/mkdumprd to a non-existent file.
96: FIPS_MODE=$(cat /proc/sys/crypto/fips_enabled)

2. For Issue 1, Make sure no kdump initial ramdisk are present in /boot folder
and start the kdump service by "service kdump start".
For issue 2, just type /sbin/mkdumprd --help.

Actual results:
Excerpt from 'service kdump start' output:
1. cat: /proc/sys/crypto/fips_enabled: No such file or directory
2. /sbin/mkdumprd: line 3923: [: : integer expression expected

Expected results:
In line 1, I think it's a good idea to let user know that fips will not be
setup instead of this message.
Line 2 is clearly not expected.

In /sbin/mkdumprd output posted above, message "cat: /proc/sys/crypto/fips_enabled: No such file or directory" must not appear.

Let me know if you need any more information.