Add Kdump option to disable reboot after save_core

By default, Kdump copies the entire vmcore to the disk and reboots the
system afterwards. While this works in most cases, it does not work for
systems hosted in RAM where the crash dump will be lost once rebooted.
To support this scenario, a new option in the kdump.conf,
save_core_to_ram, is added and commented out. Setting this option to 1
will allow a compressed crash dump to be produced by utilizing the
core_collector option if specified, otherwise the dmesg log will be
extracted from the vmcore. This is especially necessary because there is
no disk to save the dump to, and there is not enough space in the
pre-allocated crashkernel memory space to save an entire vmcore.

Author: nangelastro

SPECS/kexec-tools/kdump.conf

@@ -162,6 +162,18 @@
 #             to send fence_kdump notifications to.
 #             (this option is mandatory to enable fence_kdump).
 #
+# save_core_to_ram <0 | 1>
+#           - By default, the entire vmcore gets copied and the system reboots
+#             after the crash dump is saved. Enabling this option by specifying
+#             1 will instead save a compressed dump and not reboot after the
+#             dump is saved. The core_collector option, if also specified, is
+#             used to determine how the crash dump should be compressed. If the
+#             core_collector option is not set, then only the dmesg log will
+#             be saved.
+#
+#             This option is particularly useful when the operating system is
+#             running in RAM and reboots would otherwise remove the crash dump.
+#
 
 #raw /dev/vg/lv_kdump
 #ext4 /dev/vg/lv_kdump
@@ -183,3 +195,4 @@ core_collector makedumpfile -l --message-level 1 -d 31
 #dracut_args --omit-drivers "cfg80211 snd" --add-drivers "ext2 ext3"
 #fence_kdump_args -p 7410 -f auto -c 0 -i 10
 #fence_kdump_nodes node1 node2
+#save_core_to_ram 0

SPECS/kexec-tools/kdumpctl

@@ -87,6 +87,19 @@ save_core()
 	fi
 }
 
+save_core_to_ram()
+{
+	coredir="/var/crash/`date +"%Y-%m-%d-%H:%M"`"
+	mkdir -p $coredir
+	CORE_COLLECTOR=`grep ^core_collector $KDUMP_CONFIG_FILE | cut -d\  -f1 --complement`
+
+	if [[ -n $CORE_COLLECTOR ]]; then
+		$CORE_COLLECTOR /proc/vmcore $coredir/vmcore
+	else
+		makedumpfile -l --message-level 1 -d 31 /proc/vmcore $coredir/vmcore
+	fi
+}
+
 rebuild_fadump_initrd()
 {
 	local target_initrd_tmp
@@ -243,7 +256,7 @@ check_config()
 		case "$config_opt" in
 		\#* | "")
 			;;
-		raw|ext2|ext3|ext4|minix|btrfs|xfs|nfs|ssh|sshkey|path|core_collector|kdump_post|kdump_pre|extra_bins|extra_modules|failure_action|default|final_action|force_rebuild|force_no_rebuild|dracut_args|fence_kdump_args|fence_kdump_nodes)
+		raw|ext2|ext3|ext4|minix|btrfs|xfs|nfs|ssh|sshkey|path|core_collector|kdump_post|kdump_pre|extra_bins|extra_modules|failure_action|default|final_action|force_rebuild|force_no_rebuild|dracut_args|fence_kdump_args|fence_kdump_nodes|save_core_to_ram)
 			# remove inline comments after the end of a directive.
 			[ -z "$config_val" ] && {
 				echo "Invalid kdump config value for option $config_opt."
@@ -1248,8 +1261,24 @@ main ()
 	case "$1" in
 	  start)
 		if [ -s /proc/vmcore ]; then
-			save_core
-			reboot
+			local _save_core_to_ram save_core_to_ram="0"
+			_save_core_to_ram=`grep ^save_core_to_ram $KDUMP_CONFIG_FILE 2>/dev/null`
+
+			if [ $? -eq 0 ]; then
+				save_core_to_ram=`echo $_save_core_to_ram | cut -d' '  -f2`
+
+				if [ "$save_core_to_ram" != "0" ] && [ "$save_core_to_ram" != "1" ]; then
+					echo "Error: save_core_to_ram value is invalid"
+					exit 1
+				fi
+			fi
+
+			if [ "$save_core_to_ram" == "1" ]; then
+				save_core_to_ram
+			else
+				save_core
+				reboot
+			fi
 		else
 			start
 		fi

SPECS/kexec-tools/kexec-tools.signatures.json

@@ -20,14 +20,14 @@
   "kdump-lib-initramfs.sh": "ff4b63f99470ee796fd7e4a532dde241c4da02909306926a08a172457775a92a",
   "kdump-lib.sh": "3d50507626d4a92b8448c7d6604923f6f460c4cb5c8b18977381a7d5e516dfba",
   "kdump-udev-throttler": "125d538a59172f779b40ea32fea1e4eb50d849f25eb2537a48328d4401136679",
-  "kdump.conf": "c4cd6c4a2f00e7df10d939bc2f35f8b6b8a3edd71245aa3ba0d7fff53d9f42e8",
+  "kdump.conf": "2e65b6237560286f74530e104ee6556387b5c08ced883b1bb0a443d1e6d3c0cb",
   "kdump.conf.5": "bab6c7b171f76e8c6df91bb7db852815caacbc3cd3d4309a1dc58d36950440e4",
   "kdump.service": "ed8d22fafa6cfafaddcfd9fec82e3fa7fa42aa0366f69676fafa97e200b0506a",
   "kdump.sysconfig": "d03a24659c08aa4ce5a8e54edbe64e149f8c572c76c4e338d84dbdfbeab8fb63",
   "kdump.sysconfig.aarch64": "a829f3a0f88d983ccec2782b048f12a2eafca71aac4dd42acac2eaea8f7dd8d9",
   "kdump.sysconfig.i386": "1175bcc11be49841d4b8c3f9612e68bb7b2e0e8f00067ffa782fd9dba04f821d",
   "kdump.sysconfig.x86_64": "850b3be780165d0721edb8d34b2ba28aafc60b29850ff627daf04190369e359c",
-  "kdumpctl": "34976fa2ca34a455f393d2ac212a628d73ba4113b44f0e819c3cb5b1195e504b",
+  "kdumpctl": "bc2cf0374bef39baf2b4ae38c1218bab30ffec8e1e90fa57fce21cdd9d61982c",
   "kdumpctl.8": "025a39d212eed0584da7c23c1bb50d6ab797a0fa3c686ac6b80391af875dbaec",
   "kexec-kdump-howto.txt": "e783390a9aa582a2b5e7981cd3277081385340b5873a3f0633c789811876eccb",
   "kexec-tools-2.0.27.tar.xz": "38d78bcfa33a88928324b57dc81f50c1dfc279eede45e80957ca18f4e84b8187",

SPECS/kexec-tools/kexec-tools.spec

@@ -6,7 +6,7 @@
 Summary:        The kexec/kdump userspace component
 Name:           kexec-tools
 Version:        2.0.27
-Release:        7%{?dist}
+Release:        8%{?dist}
 License:        GPLv2
 Vendor:         Microsoft Corporation
 Distribution:   Azure Linux
@@ -330,6 +330,9 @@ done
 /usr/share/makedumpfile/
 
 %changelog
+* Mon Dec 13 2024 Nicholas Angelastro <nangelastro@microsoft.com> - 2.0.27-8
+- Support saving crash dumps on systems hosted in RAM
+
 * Tue Jul 09 2024 Chris Co <chrco@microsoft.com> - 2.0.27-7
 - Remove requires on dhcp-client