svn commit: r268660 - head/sys/amd64/amd64

[Ian Lepore]

On Tue, 2014-07-15 at 09:30 +0000, Konstantin Belousov wrote:
> Author: kib
> Date: Tue Jul 15 09:30:43 2014
> New Revision: 268660
> URL: http://svnweb.freebsd.org/changeset/base/268660
> 
> Log:
>   Make amd64 pmap_copy_pages() functional for pages not mapped by DMAP.
>   
>   Requested and reviewed by:	royger
>   Tested by:	pho, royger
>   Sponsored by:	The FreeBSD Foundation
>   MFC after:	1 week
> 
> Modified:
>   head/sys/amd64/amd64/pmap.c
> 
> Modified: head/sys/amd64/amd64/pmap.c
> ==============================================================================
> --- head/sys/amd64/amd64/pmap.c	Tue Jul 15 05:45:50 2014	(r268659)
> +++ head/sys/amd64/amd64/pmap.c	Tue Jul 15 09:30:43 2014	(r268660)
> @@ -390,6 +390,11 @@ SYSCTL_PROC(_vm_pmap, OID_AUTO, pcid_sav
>      CTLFLAG_MPSAFE, NULL, 0, pmap_pcid_save_cnt_proc, "QU",
>      "Count of saved TLB context on switch");
>  
> +/* pmap_copy_pages() over non-DMAP */
> +static struct mtx cpage_lock;
> +static vm_offset_t cpage_a;
> +static vm_offset_t cpage_b;
> +
>  /*
>   * Crashdump maps.
>   */
> @@ -1055,6 +1060,10 @@ pmap_init(void)
>  	    M_WAITOK | M_ZERO);
>  	for (i = 0; i < pv_npg; i++)
>  		TAILQ_INIT(&pv_table[i].pv_list);
> +
> +	mtx_init(&cpage_lock, "cpage", NULL, MTX_DEF);
> +	cpage_a = kva_alloc(PAGE_SIZE);
> +	cpage_b = kva_alloc(PAGE_SIZE);
>  }
>  
>  static SYSCTL_NODE(_vm_pmap, OID_AUTO, pde, CTLFLAG_RD, 0,
> @@ -5064,19 +5073,58 @@ pmap_copy_pages(vm_page_t ma[], vm_offse
>      vm_offset_t b_offset, int xfersize)
>  {
>  	void *a_cp, *b_cp;
> +	vm_page_t m_a, m_b;
> +	vm_paddr_t p_a, p_b;
> +	pt_entry_t *pte;
>  	vm_offset_t a_pg_offset, b_pg_offset;
>  	int cnt;
> +	boolean_t pinned;
>  
> +	pinned = FALSE;
>  	while (xfersize > 0) {
>  		a_pg_offset = a_offset & PAGE_MASK;
> -		cnt = min(xfersize, PAGE_SIZE - a_pg_offset);
> -		a_cp = (char *)PHYS_TO_DMAP(ma[a_offset >> PAGE_SHIFT]->
> -		    phys_addr) + a_pg_offset;
> +		m_a = ma[a_offset >> PAGE_SHIFT];
> +		p_a = m_a->phys_addr;
>  		b_pg_offset = b_offset & PAGE_MASK;
> +		m_b = mb[b_offset >> PAGE_SHIFT];
> +		p_b = m_b->phys_addr;
> +		cnt = min(xfersize, PAGE_SIZE - a_pg_offset);
>  		cnt = min(cnt, PAGE_SIZE - b_pg_offset);
> -		b_cp = (char *)PHYS_TO_DMAP(mb[b_offset >> PAGE_SHIFT]->
> -		    phys_addr) + b_pg_offset;
> +		if (__predict_false(p_a < DMAP_MIN_ADDRESS ||
> +		    p_a > DMAP_MIN_ADDRESS + dmaplimit)) {
> +			mtx_lock(&cpage_lock);
> +			sched_pin();
> +			pinned = TRUE;
> +			pte = vtopte(cpage_a);
> +			*pte = p_a | X86_PG_A | X86_PG_V |
> +			    pmap_cache_bits(kernel_pmap, m_a->md.pat_mode, 0);
> +			invlpg(cpage_a);
> +			a_cp = (char *)cpage_a + a_pg_offset;
> +		} else {
> +			a_cp = (char *)PHYS_TO_DMAP(p_a) + a_pg_offset;
> +		}
> +		if (__predict_false(p_b < DMAP_MIN_ADDRESS ||
> +		    p_b > DMAP_MIN_ADDRESS + dmaplimit)) {
> +			if (!pinned) {
> +				mtx_lock(&cpage_lock);
> +				sched_pin();
> +				pinned = TRUE;
> +			}
> +			pte = vtopte(cpage_b);
> +			*pte = p_b | X86_PG_A | X86_PG_M | X86_PG_RW |
> +			    X86_PG_V | pmap_cache_bits(kernel_pmap,
> +			    m_b->md.pat_mode, 0);
> +			invlpg(cpage_b);
> +			b_cp = (char *)cpage_b + b_pg_offset;
> +		} else {
> +			b_cp = (char *)PHYS_TO_DMAP(p_b) + b_pg_offset;
> +		}
>  		bcopy(a_cp, b_cp, cnt);
> +		if (__predict_false(pinned)) {
> +			sched_unpin();
> +			mtx_unlock(&cpage_lock);
> +			pinned = FALSE;

Should this pinned = FALSE be done under the cpage_lock to avoid a race?

-- Ian

> +		}
>  		a_offset += cnt;
>  		b_offset += cnt;
>  		xfersize -= cnt;
>