re: port-arm/55737: Apparent bug in evbarm64 DMA code causes filesystem corruption

[matthew green <mrg%eterna.com.au@localhost>]

matthew green writes:
> i think both are wrong.  it should start true, and if any
> entry is outside the parent range, set to false.

i meant to include this untested patch.  not meant to be
commited as-is, but should confirm or deny my guess with
a repro attempt.


.mrg.


Index: bus_dma.c
===================================================================
RCS file: /cvsroot/src/sys/arch/arm/arm32/bus_dma.c,v
retrieving revision 1.123
diff -p -u -r1.123 bus_dma.c
--- bus_dma.c	8 Sep 2020 10:30:17 -0000	1.123
+++ bus_dma.c	20 Oct 2020 06:29:52 -0000
@@ -1816,7 +1816,7 @@ _bus_dmatag_subregion(bus_dma_tag_t tag,
 
 #ifdef _ARM32_NEED_BUS_DMA_BOUNCE
 	struct arm32_dma_range *dr;
-	bool subset = false;
+	bool subset = true;
 	size_t nranges = 0;
 	size_t i;
 	for (i = 0, dr = tag->_ranges; i < tag->_nranges; i++, dr++) {
@@ -1826,7 +1826,7 @@ _bus_dmatag_subregion(bus_dma_tag_t tag,
 		 */
 		if (dr->dr_sysbase >= min_addr
 		    && dr->dr_sysbase + dr->dr_len - 1 <= max_addr) {
-			subset = true;
+			//subset = true;
 		} else {
 			subset = false;
 		}