Re: [PATCH 6.18.47-rc1 01/10] Revert "selinux: require a class's permission values to cover its permission count"

[Greg KH <gregkh@linuxfoundation.org>]

On Thu, Aug 27, 2026 at 05:32:27PM +0800, Wentao Guan wrote:
> > On Thu, Aug 27, 2026 at 05:07:59PM +0800, Wentao Guan wrote:
> > > Sorry for prev email title is garbaged...
> > > 
> > > > Why?  And this is not a real git id in the linux-stable tree.
> > > in stable-rc git tree.
> > > 
> > > > 
> > > > Do you want me to just drop all of the selinux patches from the 6.18.y
> > > > queue right now before I do the release?  I'll be glad to do so, but I
> > > > need to know why.
> > > 
> > > fair,
> > > 
> > > What happened in my side is Github Copliot review point out the issue:
> > > "This validation runs after symtab_insert(), so the table already owns key
> > > and perdatum. On an out-of-range value, the bad path frees both while leaving
> > > dangling pointers in the hash table, and rc is still 0, so malformed policy
> > > data is accepted and later accesses can use freed memory.
> > > Validate the value before inserting it and return -EINVAL."
> > 
> > I have no idea what that means at all as I can't parse LLM output well.
> > 
> > > In continues review, there LLM point out the mainline don`t had the issuse,
> > > for correct sequence is that git log --oneline
> > > (v6.18.46-1) selinux: require a class's permission values to cover its permission count
> > > selinux: reject a class permission count below its inherited common
> > > selinux: reject a permission value exceeding the class permission count
> > > selinux: more strict policy parsing
> > > selinux: use u16 for security classes
> > > Revert "selinux: reject a class permission count below its inherited common"
> > > Revert "selinux: reject a permission value exceeding the class permission count"
> > > (tag: v6.18.46, stable-gh/linux-6.18.y) Linux 6.18.46
> > > 
> > > This instead of Sasha`s series:
> > > 2026-08-24 22:00 ` [PATCH 6.18.y 1/3] selinux: use u16 for security classes Sasha Levin
> > > 2026-08-24 22:00   ` [PATCH 6.18.y 2/3] selinux: more strict policy parsing Sasha Levin
> > > 2026-08-24 22:00   ` [PATCH 6.18.y 3/3] selinux: require a class's permission values to cover its permission count Sasha Levin
> > 
> > Ok, I'll drop these for now, and wait for your backports.
> > 
> > But note, I'll drop this for 6.18 and older, so I need backports for ALL
> > of those branches.
> 
> for 6.18.y:
> 1. https://lore.kernel.org/stable/20260825100338.4039846-1-guanwentao@uniontech.com/
> 2 .https://lore.kernel.org/stable/20260825100338.4039846-2-guanwentao@uniontech.com/
> 3 .https://lore.kernel.org/stable/20260825100338.4039846-3-guanwentao@uniontech.com/
> 4 .https://lore.kernel.org/stable/20260825100338.4039846-4-guanwentao@uniontech.com/
> 5 .https://lore.kernel.org/stable/20260825100338.4039846-5-guanwentao@uniontech.com/
> 6 .https://lore.kernel.org/stable/20260825100338.4039846-6-guanwentao@uniontech.com/
> 7 .https://lore.kernel.org/stable/20260825100338.4039846-7-guanwentao@uniontech.com/
> 
> for 7.1.y:
> 1. https://lore.kernel.org/stable/20260827054709.4027356-1-guanwentao@uniontech.com/
> 2. https://lore.kernel.org/stable/20260827054709.4027356-1-guanwentao@uniontech.com/
> 3. https://lore.kernel.org/stable/20260827054709.4027356-1-guanwentao@uniontech.com/
> 4. https://lore.kernel.org/stable/20260827054709.4027356-1-guanwentao@uniontech.com/
> 5. https://lore.kernel.org/stable/20260827054709.4027356-1-guanwentao@uniontech.com/
> 
> There only different between 6.18.y and 7.1.y are 6.18 revert which applied not
> cleanly:
> 62be2361da29505d3183da854f1c98d2e208e852
> 9f721b5d9e1d956ffe965bc09375989eb51881fd
> and next reapplied in next series cleanly.

Again, I am totally confused.  I'll go drop ALL of the selinux changes
in ALL of the queues and wait for the next round of releases before I
pick up the above ones (and we need that same series for older trees
too.)

thanks,

greg k-h