diff mbox series

mm: kmemleak: alloc gray object for reserved region with direct map.

Message ID 20211123090641.3654006-1-calvinzhang.cool@gmail.com (mailing list archive)
State New
Headers show
Series mm: kmemleak: alloc gray object for reserved region with direct map. | expand

Commit Message

Calvin Zhang Nov. 23, 2021, 9:06 a.m. UTC
Reserved regions with direct mapping may contain references to other
regions. CMA region with fixed location is reserved without creating
kmemleak_object for it.

So add them as gray kmemleak objects.

Signed-off-by: Calvin Zhang <calvinzhang.cool@gmail.com>
---
 drivers/of/fdt.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

Comments

Catalin Marinas Nov. 25, 2021, 5:17 p.m. UTC | #1
On Tue, Nov 23, 2021 at 05:06:41PM +0800, Calvin Zhang wrote:
> Reserved regions with direct mapping may contain references to other
> regions. CMA region with fixed location is reserved without creating
> kmemleak_object for it.
> 
> So add them as gray kmemleak objects.

Do you get any kmemleak false positives without this patch? It would be
good to include them in the commit message.

Without seeing a false positive caused by this, I'm not convinced it is
the right approach. You mentioned CMA but telling kmemleak about the
whole CMA region is a pretty big hammer. I'd rather add individual
kmemleak_alloc_*() calls in cma_alloc().
Calvin Zhang Nov. 26, 2021, 1:35 a.m. UTC | #2
On Thu, Nov 25, 2021 at 05:17:53PM +0000, Catalin Marinas wrote:
>On Tue, Nov 23, 2021 at 05:06:41PM +0800, Calvin Zhang wrote:
>> Reserved regions with direct mapping may contain references to other
>> regions. CMA region with fixed location is reserved without creating
>> kmemleak_object for it.
>> 
>> So add them as gray kmemleak objects.
>
>Do you get any kmemleak false positives without this patch? It would be
>good to include them in the commit message.

Sorry, no. I thought it was possible before I saw this commit:
620951e27457 ("mm/cma: make kmemleak ignore CMA regions"). 

>
>Without seeing a false positive caused by this, I'm not convinced it is
>the right approach. You mentioned CMA but telling kmemleak about the
>whole CMA region is a pretty big hammer. I'd rather add individual
>kmemleak_alloc_*() calls in cma_alloc().

Yeah, I agree.

--
Calvin
diff mbox series

Patch

diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
index bdca35284ceb..116c582fea7a 100644
--- a/drivers/of/fdt.c
+++ b/drivers/of/fdt.c
@@ -26,6 +26,7 @@ 
 #include <linux/serial_core.h>
 #include <linux/sysfs.h>
 #include <linux/random.h>
+#include <linux/kmemleak.h>
 
 #include <asm/setup.h>  /* for COMMAND_LINE_SIZE */
 #include <asm/page.h>
@@ -522,9 +523,12 @@  static int __init __reserved_mem_reserve_reg(unsigned long node,
 		size = dt_mem_next_cell(dt_root_size_cells, &prop);
 
 		if (size &&
-		    early_init_dt_reserve_memory_arch(base, size, nomap) == 0)
+		    early_init_dt_reserve_memory_arch(base, size, nomap) == 0) {
 			pr_debug("Reserved memory: reserved region for node '%s': base %pa, size %lu MiB\n",
 				uname, &base, (unsigned long)(size / SZ_1M));
+			if (!nomap)
+				kmemleak_alloc_phys(base, size, 0, 0);
+		}
 		else
 			pr_info("Reserved memory: failed to reserve memory for node '%s': base %pa, size %lu MiB\n",
 				uname, &base, (unsigned long)(size / SZ_1M));