net/octeontx: fix build with SVE
authorRuifeng Wang <ruifeng.wang@arm.com>
Tue, 12 Jan 2021 02:57:06 +0000 (02:57 +0000)
committerDavid Marchand <david.marchand@redhat.com>
Thu, 14 Jan 2021 15:42:25 +0000 (16:42 +0100)
Building with gcc 10.2 with SVE extension enabled got error:

{standard input}: Assembler messages:
{standard input}:91: Error: selected processor does not support `addvl x4,x8,#-1'
{standard input}:95: Error: selected processor does not support `ptrue p1.d,all'
{standard input}:135: Error: selected processor does not support `whilelo p2.d,xzr,x5'
{standard input}:137: Error: selected processor does not support `decb x1'

This is because inline assembly code explicitly resets cpu model to
not have SVE support. Thus SVE instructions generated by compiler
auto vectorization got rejected by assembler.

Added SVE to the cpu model specified by inline assembly for SVE support.
Not replacing the inline assembly with C atomics because the driver relies
on specific LSE instruction to interface to co-processor [1].

Fixes: f0c7bb1bf778 ("net/octeontx/base: add octeontx IO operations")
Cc: stable@dpdk.org
[1] https://mails.dpdk.org/archives/dev/2021-January/196092.html

Signed-off-by: Ruifeng Wang <ruifeng.wang@arm.com>
Reviewed-by: Jerin Jacob <jerinj@marvell.com>
drivers/net/octeontx/base/octeontx_io.h

index 04b9ce1..d0b9cfb 100644 (file)
@@ -52,6 +52,11 @@ do {                                                 \
 #endif
 
 #if defined(RTE_ARCH_ARM64)
+#if defined(__ARM_FEATURE_SVE)
+#define __LSE_PREAMBLE " .cpu  generic+lse+sve\n"
+#else
+#define __LSE_PREAMBLE " .cpu  generic+lse\n"
+#endif
 /**
  * Perform an atomic fetch-and-add operation.
  */
@@ -61,7 +66,7 @@ octeontx_reg_ldadd_u64(void *addr, int64_t off)
        uint64_t old_val;
 
        __asm__ volatile(
-               " .cpu          generic+lse\n"
+               __LSE_PREAMBLE
                " ldadd %1, %0, [%2]\n"
                : "=r" (old_val) : "r" (off), "r" (addr) : "memory");
 
@@ -98,12 +103,13 @@ octeontx_reg_lmtst(void *lmtline_va, void *ioreg_va, const uint64_t cmdbuf[],
 
                /* LDEOR initiates atomic transfer to I/O device */
                __asm__ volatile(
-                       " .cpu          generic+lse\n"
+                       __LSE_PREAMBLE
                        " ldeor xzr, %0, [%1]\n"
                        : "=r" (result) : "r" (ioreg_va) : "memory");
        } while (!result);
 }
 
+#undef __LSE_PREAMBLE
 #else
 
 static inline uint64_t