On Thu, Nov 18, 2021 at 8:57 AM Eric Dumazet <edumazet@google.com> wrote:

>
> Unless fixups can be handled, the signature of the function needs to
> be different.
>
> In UM, we would need to provide a number of bytes that can be read.

We can make this a bit less ugly  of course.

diff --git a/arch/x86/lib/csum-partial_64.c b/arch/x86/lib/csum-partial_64.c
index 5ec35626945b6db2f7f41c6d46d5e422810eac46..7a3c4e7e05c4b21566e1ee3813a071509a9d54ff
100644
--- a/arch/x86/lib/csum-partial_64.c
+++ b/arch/x86/lib/csum-partial_64.c
@@ -21,6 +21,25 @@ static inline unsigned short from32to16(unsigned a)
        return b;
 }

+
+static inline unsigned long load_partial_long(const void *buff, int len)
+{
+#ifndef CONFIG_DCACHE_WORD_ACCESS
+               union {
+                       unsigned long   ulval;
+                       u8              bytes[sizeof(long)];
+               } v;
+
+               v.ulval = 0;
+               memcpy(v.bytes, buff, len);
+               return v.ulval;
+#else
+               unsigned int shift = (sizeof(long) - len) * BITS_PER_BYTE;
+
+               return (load_unaligned_zeropad(buff) << shift) >> shift;
+#endif
+}
+
 /*
  * Do a checksum on an arbitrary memory area.
  * Returns a 32bit checksum.
@@ -91,11 +110,9 @@ __wsum csum_partial(const void *buff, int len, __wsum sum)
                        : "memory");
                buff += 8;
        }
-       if (len & 7) {
-               unsigned int shift = (8 - (len & 7)) * 8;
-               unsigned long trail;
-
-               trail = (load_unaligned_zeropad(buff) << shift) >> shift;
+       len &= 7;
+       if (len) {
+               unsigned long trail = load_partial_long(buff, len);

                asm("addq %[trail],%[res]\n\t"
                    "adcq $0,%[res]"