=Comment: #0================================================= ---Problem Description--- RHEL53 Beta1: network installation through cxgb3 interface failed if the adapter firmware doesn't match the cxgb3 device driver requst firmware level in rhel53. Contact Information = wenxiong.com ---uname output--- 2.6.18-120.el5 Machine Type = eClipz L4 ---Debugger--- A debugger is not configured ---Steps to Reproduce--- none ---Installation and Packaging Component Data--- Install disk info: SAS Install method: NIM Install ISO Information: rhel53 beta1 ================================================================================= This bug is found when we verified our feature 201290. LTC bugzilla #43344. [Bug 43344] LTC:5.3:201290:Enable the Chelsio 10Gb adapter - driver cxgb3 - as a bootable/installable device The cxgb3 device driver binds with certain level adapter firmware. For example, in rhel52, cxgb3 driver binds with 5.0.0 level adapter firmware. In rhel53, cxgb3 driver binds with 6.0.0 level adapter firmware. When make cxgb3 interface up, cxgb3 device driver will check the level of adapter firmware. If the level doesn't match between the leval that cxgb3 driver request and the level on the adapter, cxgb3 interface up failed. When I verified this install/boot feature, if the adapter firmware is 6.0.0 now, I can do network installation through cxgb3 interface, no problem. The issue is: If the adapter firmware is 5.0.0, the cxgb3 driver modules during network installation requested 6.0.0 level firmware, the cxgb3 interface failed at ifconfig up. So network installation failed through cxgb3 interface. Can you add /lib/firmware/t3fw-6.0.0.bin file duing the network installation? So if the adapter firmware doesn't match with cxgb3 driver requirment during cxgb3 network installation, cxgb3 driver/modules can flash to the firmware to /lib/firmware/t3fw-6.0.0.bin -correct level in rhel53. If you have any questions, please let me know. Thanks for your help
If we include the cxgb3-firmware rpm as part of the initrd this will be solved. That firmware is not part of the supplemental disc, so there should be no problem adding this.
Andy, Thank you very much! Can we expect the fix in snapshot2?
------- Comment From daisyc.com 2008-11-06 16:36 EDT------- Red Hat, Just to clarify the regression situation. It is true that the cxgb3-firmware-level package is a new package for 5.3. It was not there and was not installed for 5.2 before. This is because the requirement of this package is a 5.3 feature - RH BZ:366861 - Ship the adapter firmware image with driver cxgb3. On RHEL5.2, the cvgb3 driver used to be able to come up and running without any problem. But with this new requirement, testers found that the cxgb3 driver cannot be configured on 5.3 Betas. This is a regression from 5.2 from a testing's perspective, which we are afraid will also be the case from customer's perspective if it is not fixed in 5.3. Thanks for linking this bug to your BZ 469774. I just raised this bug to "ship issue" just to make sure that the proposed change will be picked up in the next snapshot. Please let us know if you have any other concerns about this. Thanks. This event sent from IssueTracker by jkachuck issue 234962
Thanks for the feedback. We definitely recognize this as an issue that we need to fix and I think the proposed change will help keep the kernel and firmware in sync between releases. New systems that come from the manufacturer might have the new firmware and would not see this issue, but systems that are older and being installed with 5.3 cleanly or ones that are being updated from 5.2 really need to make sure they are getting the right firmware. I think our change will make this all work correctly.
*** Bug 43344 has been marked as a duplicate of this bug. ***
My test kernels have been updated to include a patch for this bugzilla. http://people.redhat.com/agospoda/#rhel5 Please test them and report back your results.
Andy, Wendy has tested the test kernel and she put her response in RIT234692 (IBM BZ 49507) Hi Andy, I installed 2.6.18-122.el5.gtest.60.ppc64.rpm and reboot with -122.el5.gtest.60 kernel. After reboot, I didn't see /lib/firmware/t3fw-6.0.0.bin file. Which package you include /lib/firmware/t3fw-6.0.0.bin in this -122.el5.gtest.60 rpm? #rpm -ql 2.6.18-122.el5.gtest.60 /boot/System.map-2.6.18-122.el5.gtest.60 /boot/config-2.6.18-122.el5.gtest.60 /boot/initrd-2.6.18-122.el5.gtest.60.img /boot/symvers-2.6.18-122.el5.gtest.60.gz /boot/vmlinuz-2.6.18-122.el5.gtest.60 /etc/modprobe.d/blacklist-firewire /lib/modules/2.6.18-122.el5.gtest.60 /lib/modules/2.6.18-122.el5.gtest.60/build /lib/modules/2.6.18-122.el5.gtest.60/extra /lib/modules/2.6.18-122.el5.gtest.60/kernel /lib/modules/2.6.18-122.el5.gtest.60/kernel/arch /lib/modules/2.6.18-122.el5.gtest.60/kernel/arch/powerpc /lib/modules/2.6.18-122.el5.gtest.60/kernel/arch/powerpc/oprofile Will attach initrd outout. Is something wrong with my test? Thanks, Wendy Andy, Please also see the initrd output that Wendy attached under RIT234692. It should contain more information for you to diagnose. Thanks for your help.
Wendy, how did you install the kernel rpm? Did you use --force or --nodeps to install the kernel? It should not have installed unless you have the cxgb3-firmware rpm installed on your system. The point of this is that: 1. When upgrading yum will note that it is needed when updating and make sure to download that package too. 2. When composing an initrd the cxgb3-firmware rpm will get added to the initrd to satisfy the kernels requirement.
I just confirmed that my test kernels cannot be installed from the command line (rpm -i) without the proper version of cxgb3-firmware rpm installed. The conflicts line added to the rpm spec-file are what makes this work properly. If you installed with --force and/or --nodeps this check would have been skipped and I suspect your cxgb3 card would still not work.
>>Wendy, how did you install the kernel rpm? Did you use --force or -->>nodeps to >>install the kernel? It should not have installed unless you have the >>cxgb3-firmware rpm installed on your system. The point of this is that: Hi Andy, #rpm -ivh kernel-2.6.18-122.el5.gtest.60.ppc64.rpm I didn't use --force or --nodeps. Also I didn't have cxgb3-firmware rpm installed on my system. Thanks, Wendy
That's interesting, because this is what I get: # rpm -ivh kernel-2.6.18-122.el5.gtest.60.x86_64.rpm error: Failed dependencies: cxgb3-firmware < 6.0.0-6 conflicts with kernel-2.6.18-122.el5.gtest.60.x86_64 Is this some ppc64 oddity? (That doesn't seem logical to me.)
>>I just confirmed that my test kernels cannot be installed from the >>command line >>(rpm -i) without the proper version of cxgb3-firmware rpm installed. >>The >>conflicts line added to the rpm spec-file are what makes this work >>properly. >>If you installed with --force and/or --nodeps this check would have >>been >>skipped and I suspect your cxgb3 card would still not work. I tried again, here is the detail commands I just used. Is your test kernel is kernel-2.6.18-122.el5.gtest.60.rpm? First, check if cxgb3-firmware and kernel-2.6.18-122.el5.gtest.60.rpm in the system. [root@IO79 ~]# uname -r 2.6.18-120.el5 [root@IO79 firmware]# rpm -qa|grep cxgb3 libcxgb3-1.2.2-1.el5 libcxgb3-1.2.2-1.el5 [root@IO79 firmware]# rpm -qa|grep 122 kernel-2.6.18-122.el5 Second, [root@IO79 ~]# rpm -ivh kernel-2.6.18-122.el5.gtest.60.ppc64.rpm Preparing... ########################################### [100%] 1:kernel ########################################### [100%] Third, check again, [root@IO79 firmware]# rpm -qa|grep cxgb3 libcxgb3-1.2.2-1.el5 libcxgb3-1.2.2-1.el5 [root@IO79 firmware]# rpm -qa|grep 122 kernel-2.6.18-122.el5 kernel-2.6.18-122.el5.gtest.60 I didn't see cxgb3-firmware installed. Am I missing something?
OK, someone just re-explained the difference between requires and conflicts in rpm spec-files, so I'll have to re-do this. This should hopefully still make snap3.
Thanks! Let me know when you have your kernel ready for me to test. Just confirm with you: The fixes you make I expect have : (1)cxgb3-firmware is automatically installed after install rhel53. (2)we can do network installation through cxgb3 interface(if the firmware on adapter doesn't match with the cxgb3 device requested firmware). Thanks for your hard work! We want to make the fixes into Snapshot3. We understood that it is hard for us to make the fixes in after SnapShot3. Wendy
Andy, Do you have test kernel around somewhere for this bug? Thanks!
Due to this bug happened during network installation, Can you add t3fw-6.0.0.bin on the install initrd or in ramdisk.image? If you can build it into next snapshot, we can give it try. Thanks, Wendy
Unfortunately it's not that easy. The installer doesn't have a way to respond appropriately to the request_firmware action, so we can't just put the files in /lib/firmware. We are working up some ways to make this happen and hope to have something soon. We will let you know.
in kernel-2.6.18-125.el5 You can download this test kernel from http://people.redhat.com/dzickus/el5
*Sigh* this patch seems to be causing problems, but I know what's wrong. :-/
Andy, I tested this new kernel with ifconfig cxgb3 interface without /lib/firmware/t3fw-6.0.0.bin, the adapter flashed to embedded firmware. Does this mean it will work in/during network installation?
Thanks for testing! I'm glad to hear that you get the same results that I get when I've been trying it. The cxgb3-based devices should now work just fine during network installation of 5.3. We will need to you to test it for sure (to be sure the short delay to load the firmware isn't a problem for anaconda), but I don't imagine it will be any problem.
in kernel-2.6.18-126.el5 You can download this test kernel from http://people.redhat.com/dzickus/el5
Andy has posted another cleanup patch. :-) Reverting back to POST to pick it up.
in kernel-2.6.18-127.el5 You can download this test kernel from http://people.redhat.com/dzickus/el5
I have tried this in snap6. The adapter has 7.0.0 level firmware, when I picked cxgb3 interface as interface during network installation, the cxgb3 driver is loaded and adapter firmware flashed to 6.0.0 level. Looks the fixes have been in snap6. Why posted -127 kernel again? Let us know if you make new changes in -127 for cxgb3 driver. Thanks, Wendy
Just a small patch that has a 2-line bug-fix in an error path. Nothing that should effect you if you system is working now, but something that's good to have.
*** Bug 50647 has been marked as a duplicate of this bug. ***
Andy, Thanks for letting us know!
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHSA-2009-0225.html