git: f3ea417f96b0 - main - x86 busdma_bounce: use malloc_domainset_aligned(9).

Konstantin Belousov kib at FreeBSD.org
Sun Jan 17 17:29:55 UTC 2021


The branch main has been updated by kib:

URL: https://cgit.FreeBSD.org/src/commit/?id=f3ea417f96b011a7eb4f43e3142e572833287ef4

commit f3ea417f96b011a7eb4f43e3142e572833287ef4
Author:     Konstantin Belousov <kib at FreeBSD.org>
AuthorDate: 2021-01-14 04:02:21 +0000
Commit:     Konstantin Belousov <kib at FreeBSD.org>
CommitDate: 2021-01-17 17:29:05 +0000

    x86 busdma_bounce: use malloc_domainset_aligned(9).
    
    This stops busdma bounce making assumptions about alignment of malloc(9)
    results, which are no longer true.
    
    Also add assert that the result of malloc_aligned() fits into single
    page, which is the assumption of the code.
    
    Reported by:    dim
    Reviewed by:    andrew, jah, markj
    Tested by:      pho
    MFC after:      1 week
    Sponsored by:   The FreeBSD Foundation
    Differential Revision:  https://reviews.freebsd.org/D28147
---
 sys/x86/x86/busdma_bounce.c | 20 ++++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

diff --git a/sys/x86/x86/busdma_bounce.c b/sys/x86/x86/busdma_bounce.c
index 65fbbb7ec1a4..75fdf4143b9b 100644
--- a/sys/x86/x86/busdma_bounce.c
+++ b/sys/x86/x86/busdma_bounce.c
@@ -433,9 +433,9 @@ bounce_bus_dmamem_alloc(bus_dma_tag_t dmat, void **vaddr, int flags,
 
 	/*
 	 * Allocate the buffer from the malloc(9) allocator if...
-	 *  - It's small enough to fit into a single power of two sized bucket.
-	 *  - The alignment is less than or equal to the maximum size
+	 *  - It's small enough to fit into a single page.
 	 *  - The low address requirement is fulfilled.
+	 *  - Default cache attributes are requested (WB).
 	 * else allocate non-contiguous pages if...
 	 *  - The page count that could get allocated doesn't exceed
 	 *    nsegments also when the maximum segment size is less
@@ -445,19 +445,19 @@ bounce_bus_dmamem_alloc(bus_dma_tag_t dmat, void **vaddr, int flags,
 	 * else allocate a block of contiguous pages because one or more of the
 	 * constraints is something that only the contig allocator can fulfill.
 	 *
-	 * NOTE: The (dmat->common.alignment <= dmat->maxsize) check
-	 * below is just a quick hack. The exact alignment guarantees
-	 * of malloc(9) need to be nailed down, and the code below
-	 * should be rewritten to take that into account.
-	 *
-	 * In the meantime warn the user if malloc gets it wrong.
+	 * Warn the user if malloc gets it wrong.
 	 */
 	if (dmat->common.maxsize <= PAGE_SIZE &&
-	    dmat->common.alignment <= dmat->common.maxsize &&
 	    dmat->common.lowaddr >= ptoa((vm_paddr_t)Maxmem) &&
 	    attr == VM_MEMATTR_DEFAULT) {
-		*vaddr = malloc_domainset(dmat->common.maxsize, M_DEVBUF,
+		*vaddr = malloc_domainset_aligned(dmat->common.maxsize,
+		    dmat->common.alignment, M_DEVBUF,
 		    DOMAINSET_PREF(dmat->common.domain), mflags);
+		KASSERT(*vaddr == NULL || ((uintptr_t)*vaddr & PAGE_MASK) +
+		    dmat->common.maxsize <= PAGE_SIZE,
+		    ("bounce_bus_dmamem_alloc: multi-page alloc %p maxsize "
+		    "%#jx align %#jx", *vaddr, (uintmax_t)dmat->common.maxsize,
+		    (uintmax_t)dmat->common.alignment));
 	} else if (dmat->common.nsegments >=
 	    howmany(dmat->common.maxsize, MIN(dmat->common.maxsegsz,
 	    PAGE_SIZE)) &&


More information about the dev-commits-src-all mailing list