crashkernel: Maximize the reserved size in osbuild case

At present, the deduction of the proper crashkernel value depends on
detecting platform details. However, more and more platforms are now
deployed using OSTree images, which means kdump-utils cannot retrieve
complete platform information. This often leads to an underestimation of
the memory required by the kdump kernel, increasing the risk of
out-of-memory (OOM) issues.

To mitigate this situation, we choose to maximize the reserved
crashkernel size in the osbuild case, and recommend users to update the
kernel command line and release the excess reserved memory after reboot
(addressed in the next patch).

Signed-off-by: Pingfan Liu <[email protected]>

Author: Pingfan Liu

kdump-lib.sh

@@ -16,6 +16,7 @@ FADUMP_REGISTER_SYS_NODE="/sys/kernel/fadump/registered"
 FADUMP_APPEND_ARGS_SYS_NODE="/sys/kernel/fadump/bootargs_append"
 # shellcheck disable=SC2034
 FENCE_KDUMP_CONFIG_FILE="/etc/sysconfig/fence_kdump"
+maximize_crashkernel=false
 
 # read the value of an environ variable from given environ file path
 #
@@ -65,7 +66,7 @@ is_aws_aarch64()
 
 is_sme_or_sev_active()
 {
-	journalctl -q --dmesg --grep "^Memory Encryption Features active: AMD (SME|SEV)$" > /dev/null 2>&1
+	$maximize_crashkernel || journalctl -q --dmesg --grep "^Memory Encryption Features active: AMD (SME|SEV)$" > /dev/null 2>&1
 }
 
 has_command()
@@ -873,12 +874,19 @@ get_recommend_size()
 
 has_mlx5()
 {
-	[[ -d /sys/bus/pci/drivers/mlx5_core ]]
+	$maximize_crashkernel || [[ -d /sys/bus/pci/drivers/mlx5_core ]]
 }
 
 has_aarch64_smmu()
 {
-	ls /sys/devices/platform/arm-smmu-* 1> /dev/null 2>&1
+	$maximize_crashkernel || ls /sys/devices/platform/arm-smmu-* 1> /dev/null 2>&1
+}
+
+is_aarch64_64k_kernel()
+{
+	local _kernel="$1"
+	# the naming convention of 64k variant suffixes with +64k, e.g. "vmlinuz-5.14.0-312.el9.aarch64+64k"
+	$maximize_crashkernel || echo "$_kernel" | grep -q 64k
 }
 
 is_memsize()
@@ -1033,6 +1041,9 @@ kdump_get_arch_recommend_crashkernel()
 	local _delta=0
 	local _skip=0
 
+	# osbuild deploys rpm on isolated environment. kdump-utils has no opportunity
+	# to deduce the exact memory cost on the real target.
+	_is_osbuild && maximize_crashkernel=true
 	if [[ -z $1 ]]; then
 		if is_fadump_capable; then
 			_dump_mode=fadump
@@ -1061,8 +1072,7 @@ kdump_get_arch_recommend_crashkernel()
 		# skip adding additional memory for small-memory machine
 		_skip=1
 
-		# the naming convention of 64k variant suffixes with +64k, e.g. "vmlinuz-5.14.0-312.el9.aarch64+64k"
-		if echo "$_running_kernel" | grep -q 64k; then
+		if is_aarch64_64k_kernel "$_running_kernel"; then
 			# Without smmu, the diff of MemFree between 4K and 64K measured on a high end aarch64 machine is 82M.
 			# Picking up 100M to cover this diff. And finally, we have "2G-4G:356M;4G-64G:420M;64G-:676M"
 			((_delta += 100))