rpms/kernel/devel neuter_intel_microcode_load.patch, NONE, 1.1 kernel.spec, 1.1916, 1.1917

Kyle McMartin kyle at fedoraproject.org
Sat Feb 6 22:41:05 UTC 2010


Author: kyle

Update of /cvs/pkgs/rpms/kernel/devel
In directory cvs1.fedora.phx.redhat.com:/tmp/cvs-serv13617

Modified Files:
	kernel.spec 
Added Files:
	neuter_intel_microcode_load.patch 
Log Message:
* Sat Feb 06 2010 Kyle McMartin <kyle at redhat.com>
- Hack around delay loading microcode.ko, on intel, we don't split out
  the firmware into cpuid specific versions (in fact, I don't know who does...)
  so just patch out the request_firmware calls in microcode_intel.c, and
  microcode_ctl.init will do the right thing. (fixes rhbz#560031)
  (side note: I'll fix microcode_ctl to do one better at some point.)


neuter_intel_microcode_load.patch:
 microcode_intel.c |   11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

--- NEW FILE neuter_intel_microcode_load.patch ---
diff --git a/arch/x86/kernel/microcode_intel.c b/arch/x86/kernel/microcode_intel.c
index ebd193e..c57024a 100644
--- a/arch/x86/kernel/microcode_intel.c
+++ b/arch/x86/kernel/microcode_intel.c
@@ -453,9 +453,18 @@ static void microcode_fini_cpu(int cpu)
 	uci->mc = NULL;
 }
 
+/* we don't ship the broken out files...
+ *  instead, we'll just fail here, and load it with microcode_ctl
+ */
+static enum ucode_state noop_request_microcode_fw(int cpu,
+	struct device *device)
+{
+	return UCODE_NFOUND;
+}
+
 static struct microcode_ops microcode_intel_ops = {
 	.request_microcode_user		  = request_microcode_user,
-	.request_microcode_fw             = request_microcode_fw,
+	.request_microcode_fw             = noop_request_microcode_fw,
 	.collect_cpu_info                 = collect_cpu_info,
 	.apply_microcode                  = apply_microcode,
 	.microcode_fini_cpu               = microcode_fini_cpu,


Index: kernel.spec
===================================================================
RCS file: /cvs/pkgs/rpms/kernel/devel/kernel.spec,v
retrieving revision 1.1916
retrieving revision 1.1917
diff -u -p -r1.1916 -r1.1917
--- kernel.spec	6 Feb 2010 17:10:15 -0000	1.1916
+++ kernel.spec	6 Feb 2010 22:41:03 -0000	1.1917
@@ -723,6 +723,8 @@ Patch12015: add-appleir-usb-driver.patch
 Patch12016: fix-conntrack-bug-with-namespaces.patch
 Patch12017: prevent-runtime-conntrack-changes.patch
 
+Patch12018: neuter_intel_microcode_load.patch
+
 %endif
 
 BuildRoot: %{_tmppath}/kernel-%{KVERREL}-root
@@ -1328,6 +1330,8 @@ ApplyPatch add-appleir-usb-driver.patch
 
 ApplyPatch fix-conntrack-bug-with-namespaces.patch
 
+ApplyPatch neuter_intel_microcode_load.patch
+
 # END OF PATCH APPLICATIONS
 
 %endif
@@ -1980,6 +1984,13 @@ fi
 
 %changelog
 * Sat Feb 06 2010 Kyle McMartin <kyle at redhat.com>
+- Hack around delay loading microcode.ko, on intel, we don't split out
+  the firmware into cpuid specific versions (in fact, I don't know who does...)
+  so just patch out the request_firmware calls in microcode_intel.c, and
+  microcode_ctl.init will do the right thing. (fixes rhbz#560031)
+  (side note: I'll fix microcode_ctl to do one better at some point.)
+
+* Sat Feb 06 2010 Kyle McMartin <kyle at redhat.com>
 - Don't want linux-firmware if %with_firmware, yet. (Think F-11/F-12 2.6.33)
 
 * Fri Feb 05 2010 Peter Jones <pjones at redhat.com>



More information about the scm-commits mailing list