Hi,
Le 19/01/2026 à 12:12, Andre Przywara a écrit :
Hi,

On 28/10/2025 09:12, Richard Genoud wrote:
Fix pointer from interget warning when compiling for ARM64

When compiling for arm64, we get this error:
error: passing argument 2 of ‘__memcpy_fromio’ makes pointer from
            integer without a cast [-Wint-conversion]

Moreover the copy should be made with dedicated readl(), like for any
register access on this peripheral, since they are 32bit wide.

So, instead of memcpy_fromio(), just use a readl() loop.
Introduce nand_readlcpy() to implement this loop.

Fixes: 6ddbb1e936c7 ("spl: nand: sunxi: use PIO instead of DMA")
Suggested-by: Andre Przywara <[email protected]>
Signed-off-by: Richard Genoud <[email protected]>
---
  drivers/mtd/nand/raw/sunxi_nand_spl.c | 14 +++++++++++++-
  1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/drivers/mtd/nand/raw/sunxi_nand_spl.c b/drivers/mtd/nand/ raw/sunxi_nand_spl.c
index 4f1e2d9a5775..a15f54573df2 100644
--- a/drivers/mtd/nand/raw/sunxi_nand_spl.c
+++ b/drivers/mtd/nand/raw/sunxi_nand_spl.c
@@ -251,6 +251,17 @@ static int nand_change_column(u16 column)
  static const int ecc_bytes[] = {32, 46, 54, 60, 74, 88, 102, 110, 116};
+static void nand_readlcpy(u32 *dest, u32 * __iomem src, size_t len)
+{
+    if (len & 0x3)
+        printf("length should be multiple of 4 (32bits access), data will be incomplete.\n");

This should not be a message printed on the console. There is only one(?) internal user, and the argument is a multiple of 512, if I see that correctly. So it's an internal error that wouldn't happen at all in the moment. It just blows up the code size, and isn't even very descriptive.

If you *really* want to check that, you could try to use something like builtin_constant_p, to create a compile-time error, or maybe use debug(). But I would suggest to drop it completely, as it would be pointless with the current code anyway.

The rest looks fine, thanks for adding that function.
Indeed, I can drop it and add a comment to warn future users in this driver.

@Michael, do you want me to do a v3, or just apply this v2 and I'll send a fix? (or you just remove those lines when you apply the patch?)

Thanks !

Regards,
Richard

Cheers,
Andre

+    len >>= 2;
+
+    while (len--)
+        *dest++ = readl(src++);
+}
+
  static int nand_read_page(const struct nfc_config *conf, u32 offs,
                void *dest, int len)
  {
@@ -310,7 +321,8 @@ static int nand_read_page(const struct nfc_config *conf, u32 offs,
              return 1;
          /* Retrieve the data from SRAM */
-        memcpy_fromio(data, SUNXI_NFC_BASE + NFC_RAM0_BASE,
+        nand_readlcpy((u32 *)data,
+                  (void *)(uintptr_t)SUNXI_NFC_BASE + NFC_RAM0_BASE,
                    conf->ecc_size);
          /* Stop the ECC engine */



--
Richard Genoud, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

Reply via email to