Thread (9 messages) 9 messages, 4 authors, 2026-03-06

Re: [PATCH 1/2] arm_mpam: Force __iomem casts

From: Ben Horgan <ben.horgan@arm.com>
Date: 2026-02-27 14:06:30
Also in: lkml
Subsystem: memory management - core, memory management - page allocator, the rest · Maintainers: Andrew Morton, David Hildenbrand, Vlastimil Babka, Linus Torvalds

Hi Krzysztof,

On 2/16/26 11:02, Krzysztof Kozlowski wrote:
quoted hunk ↗ jump to hunk
Code allocates standard kernel memory to pass to the MPAM, which expects
__iomem.  The code is safe, because __iomem accessors should work fine
on kernel mapped memory, however leads to sparse warnings:

  test_mpam_devices.c:327:42: warning: incorrect type in initializer (different address spaces)
  test_mpam_devices.c:327:42:    expected char [noderef] __iomem *buf
  test_mpam_devices.c:327:42:    got void *
  test_mpam_devices.c:342:24: warning: cast removes address space '__iomem' of expression

Cast the pointer to memory via __force to silence them.

Signed-off-by: Krzysztof Kozlowski <redacted>
---
 drivers/resctrl/test_mpam_devices.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/resctrl/test_mpam_devices.c b/drivers/resctrl/test_mpam_devices.c
index 3e8d564a0c64..2de41b47c138 100644
--- a/drivers/resctrl/test_mpam_devices.c
+++ b/drivers/resctrl/test_mpam_devices.c
@@ -324,7 +324,7 @@ static void test_mpam_enable_merge_features(struct kunit *test)
 
 static void test_mpam_reset_msc_bitmap(struct kunit *test)
 {
-	char __iomem *buf = kunit_kzalloc(test, SZ_16K, GFP_KERNEL);
+	char __iomem *buf = (__force char __iomem *)kunit_kzalloc(test, SZ_16K, GFP_KERNEL);
 	struct mpam_msc fake_msc = {};
 	u32 *test_result;
 
@@ -339,7 +339,7 @@ static void test_mpam_reset_msc_bitmap(struct kunit *test)
 	mutex_init(&fake_msc.part_sel_lock);
 	mutex_lock(&fake_msc.part_sel_lock);
 
-	test_result = (u32 *)(buf + MPAMCFG_CPBM);
+	test_result = (__force u32 *)(buf + MPAMCFG_CPBM);
 
 	mpam_reset_msc_bitmap(&fake_msc, MPAMCFG_CPBM, 0);
 	KUNIT_EXPECT_EQ(test, test_result[0], 0);

This change looks good to me. As sparse is more broken I needed to use
the patch from [1] to reproduce this. Copied here for convenience.
diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index 2b30a0529d48..90536b2bc42e 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -14,8 +14,8 @@ struct vm_area_struct;
 struct mempolicy;

 /* Helper macro to avoid gfp flags if they are the default one */
-#define __default_gfp(a,...) a
-#define default_gfp(...) __default_gfp(__VA_ARGS__ __VA_OPT__(,)
GFP_KERNEL)
+#define __default_gfp(a,b,...) b
+#define default_gfp(...) __default_gfp(,##__VA_ARGS__,GFP_KERNEL)

 /* Convert GFP flags to their corresponding migrate type */
 #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)

There is a kernel test robot report [2] and that asks for these tags:

Reported-by: kernel test robot <redacted>
Closes:
https://lore.kernel.org/oe-kbuild-all/202512160133.eAzPdJv2-lkp@intel.com/ (local)

Acked-by: Ben Horgan <ben.horgan@arm.com>

[1]
https://lore.kernel.org/all/CAHk-=wijD-giccF6sJ+BdJpGDX9kPEUT6kryaQG0GRyJ3QQwng@mail.gmail.com/
[2] https://lore.kernel.org/all/202512160133.eAzPdJv2-lkp@intel.com/ (local)

Thanks,

Ben

Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help