Patches potentially missing from stable releases

We did some work on analyzing patches potentially missing from stable releases based on the Fixes: and Revert references in the commit messages. Our script is based on similar idea as described by Guenter Roeck in this earlier mail:
https://lore.kernel.org/stable/[email protected]/.

Although the list is not comprehensive, we figured it makes sense to publish it in case the results are of interest to someone else also.

The below list of potentially missing commits is based on 4.19, but some of the commits might also apply to 5.4 and 5.6.

For each potentially missing commit flagged by the script, we read the commit message and had a short look at the change. We then added comments on our own judgement if it might be stable material or not. No comments simply means the potentially missing patch appears stable material. "Based on commit" is the mainline patch that has been backported to 4.19 and is referenced by the missing commit. We did not check if the patch applies without changes, nor did we build or execute any tests.

6011002c1584 uio: make symbol 'uio_class_registered' static

    Based on commit: ae61cf5b9913

968339fad422 powerpc/boot: Delete unneeded .globl _zimage_start
Based on commit: ee9d21b3b358
Comment: not stable material?

ed4d81c4b3f2 net: aquantia: when cleaning hw cache it should be toggled
Based on commit: 7a1bb49461b1
Comment: this was backported to 5.3, but for some reason not to older stable releases

b27507bb59ed net/ibmvnic: unlock rtnl_lock in reset so linkwatch_event can ru

    Based on commit: a5681e20b541

61c347ba5511 afs: Clear AFS_VNODE_CB_PROMISED if we detect callback expiry
Based on commit: ae3b7361dc0e
Comment: likely requires manual backport

1a49b2fd8f58 kbuild: strip whitespace in cmd_record_mcount findstring
Based on commit: 6977f95e63b9
Comment: not stable material?

669e859b5ea7 btrfs: drop the lock on error in btrfs_dev_replace_cancel

    Based on commit: d189dd70e255
    Comment: earlier backport failed, this would likely require manual 
    backport: https://lore.kernel.org/stable/[email protected]/

dddaf89e2fbc netfilter: ipt_CLUSTERIP: make symbol 'cip_netdev_notifier' stat

    Based on commit: 5a86d68bcf02

01091c496f92 acpi/nfit: improve bounds checking for 'func'
Based on commit: 11189c1089da
Comment: the missing commit was picked by AUTOSEL to 5.4, 5.5, and 5.6, but for some reason, it was not backported 4.19:
https://lore.kernel.org/stable/?q=01091c496f92*

2b74c878b0ea IB/hfi1: Unreserve a flushed OPFN request

    Based on commit: ca95f802ef51
    Comment: earlier backport failed, this would likely require manual 
    backport: https://lore.kernel.org/stable/[email protected]/

0b815023a1d4 bnxt_en: Fix ring checking logic on 57500 chips.
Based on commit: 36d65be9a880
Comment: likely requires manual backport

6ae865615fc4 x86/uaccess: Dont leak the AC flag into __put_user() argument ev

    Based on commit: 2a418cf3f5f1
    Comment: commit 9b8bd476e78e which mentions complementing 
    6ae865615fc4, has already been backported to 4.19

70db5b04cbe1 f2fs: give some messages for inline_xattr_size
Based on commit: 500e0b28ecd3
Comment: not stable material?

2a06b8982f8f net: reorder 'struct net' fields to avoid false sharing
Based on commit: 355b98553789
Comment: this was backported to 5.3, but for some reason not to older stable releases

0fbf21c3b36a ALSA: hda/realtek - Enable micmute LED for Huawei laptops
Based on commit: 8ac51bbc4cfe
Comment: not stable material?

Thanks,
-- Henri

That last sentence should have been a huge red flag when writing it and sending out this email...

You looked at this patch and thought it was relevant for backporting?

Why? It "obviously" does not fix anything, and is just a cleanup patch.

Why did it pass your criteria?

Ok, I'm going to stop here.

I appreciate sending lists of commits that you have determined should be applied, and will be glad to review them, but you don't have to give me a list of all potential patches and your comments on them, as that doesn't help much.

I have scripts that can churn out the false-positive lists like these, that's relatively easy to create. It's the curated lists that you have looked at and reviewed and determined, in your opinion, should be applied that are much more valuable and better to work with.

At this point, I will need others to do backporting for older kernels like this, unless there is a good reason why you can't do it yourself. That shows you actually want the patch to be backported, as you have done the effort to do so and check that it builds properly.

Again, random lists aren't all that helpful, but curated ones are.

This was known not to work. I asked for help, and just asking for help again isn't probably going to do much :(

It doesn't apply and needs manual backporting. Why didn't you test that?

Again, curated ids, and backported patches are the best thing you can do to help out here. See Guenter's email for an example of the first, and Ben's emails of backported patches for examples of the second.

thanks,

greg k-h

Thank you for your comments.

This list was a byproduct of other analysis we did, I did not intent to ask these patches for inclusion into stable. We simply wanted to report this list in case it includes some patches that might have dropped off your radar. We are aware that such information without further work, e.g. backporting and testing does not add much yet.

Thanks,
-- Henri