Hi Peter,
On 2024/12/11 21:34, Peter Zijlstra wrote:
On Wed, Dec 11, 2024 at 01:56:30PM +0100, Peter Zijlstra wrote:
On Wed, Dec 11, 2024 at 06:31:31PM +0800, Qi Zheng wrote:
Now, it is also possible not to use the mmu gather mechanism to
free PTE pages:
pte_free_tlb
--> ___pte_free_tlb
--> pagetable_pte_dtor
--> ptlock_free
paravirt_tlb_remove_table
--> free PTE page
pte_free
--> pagetable_pte_dtor
--> ptlock_free
pagetable_free
If we want to move the freeing of the ptlock into __tlb_remove_table(),
we may need to define a pagetable_pte_dtor_no_ptlock() and let
___pte_free_tlb() to call it.
And ALLOC_SPLIT_PTLOCKS only be enabled when the spinlock debug function
is turned on, so I finally choose to only change ptlock itself without
changing the general path.
PREEMPT_RT will also have it on, and it feels a bit odd to have two
disparate RCU frees for something so very closely related.
Got it.
We could push the whole of pagetable_*_dtor into tlb_remove_table(),
except that seems to cause trouble for p4d, which is not conforming.
Argh, all this is a giant trainwreck and needs some serious cleanup, and
Indeed.
I feel your patch is just making it worse.
:(
Notably:
- s390 pud isn't calling the existing pagetable_pud_[cd]tor()
- none of the p4d things have pagetable_p4d_[cd]tor() (x86,arm64,s390,riscv)
and they have inconsistent accounting
- while much of the _ctor calls are in generic code, many of the _dtor
calls are in arch code for hysterial raisins, this could easily be
fixed
- if we fix ptlock_free() to handle NULL, then all the _dtor()
functions can use it, and we can observe they're all identical
and can be folded
after all that cleanup, you can move the _dtor from *_free_tlb() into
tlb_remove_table() -- which for the above case, would then have it
called from __tlb_remove_table_free().
Thomas, you were looking for technical debt -- there's a ton of that
around here :/
Some of the above are in the below... I gave up after a while.
Thank you for such a detail suggestion! I will try to digest it and
implement it.
Thanks!
