On Sun, Nov 09, 2025 at 10:20:02AM +0000, Martin Pieuchot wrote:
> uvm_pagermapin() is called with the WAITOK flag during page faults in
> uvn_io().  There, it is expected to block if necessary and return a
> valid kva.
> 
> bluhm@ reported in august a bug where this didn't happen:
>   https://marc.info/?l=openbsd-bugs&m=175648751931652&w=2
> 
> Diff below fixes the issue by waking the pdaemon and sleeping for memory
> if pmap_enter(9) fails.
> 
> Note that the pdaemon doesn't pass the WAITOK flag when it calls this
> function.

I doubt I have really exercised this code path on these machines, but
this went through a few bulk builds on arm64 and sparc64.  Anyway,
waiting here makes more sense than crashing.

> ok?

ok jca@

> Index: uvm/uvm_pager.c
> ===================================================================
> RCS file: /cvs/src/sys/uvm/uvm_pager.c,v
> diff -u -p -r1.94 uvm_pager.c
> --- uvm/uvm_pager.c	10 Mar 2025 14:13:58 -0000	1.94
> +++ uvm/uvm_pager.c	6 Oct 2025 08:13:07 -0000
> @@ -263,13 +263,16 @@ uvm_pagermapin(struct vm_page **pps, int
>  		pp = *pps++;
>  		KASSERT(pp);
>  		KASSERT(pp->pg_flags & PG_BUSY);
> -		/* Allow pmap_enter to fail. */
> -		if (pmap_enter(pmap_kernel(), cva, VM_PAGE_TO_PHYS(pp),
> +		while (pmap_enter(pmap_kernel(), cva, VM_PAGE_TO_PHYS(pp),
>  		    prot, PMAP_WIRED | PMAP_CANFAIL | prot) != 0) {
> -			pmap_remove(pmap_kernel(), kva, cva);
> -			pmap_update(pmap_kernel());
> -			uvm_pseg_release(kva);
> -			return 0;
> +		    	if (flags & UVMPAGER_MAPIN_WAITOK)
> +		    		uvm_wait("pgrmapin");
> +			else {
> +				pmap_remove(pmap_kernel(), kva, cva);
> +				pmap_update(pmap_kernel());
> +				uvm_pseg_release(kva);
> +				return 0;
> +			}
>  		}
>  	}
>  	pmap_update(pmap_kernel());
> 
> 

-- 
jca