[RISCV] Add partial validation of S and X extension names to RISCVISAInfo::parseNormalizedArchString.

Extensions starting with 's' or 'x' should always be followed by an
alphabetical character.  I don't know of any crashes from this currently,
but it seemed better to be defensive.

Author: topperc

llvm/lib/TargetParser/RISCVISAInfo.cpp

@@ -486,9 +486,11 @@ RISCVISAInfo::parseNormalizedArchString(StringRef Arch) {
       return createStringError(errc::invalid_argument,
                                "failed to parse major version number");
 
-    if (ExtName[0] == 'z' && (ExtName.size() == 1 || isDigit(ExtName[1])))
+    if ((ExtName[0] == 'z' || ExtName[0] == 's' || ExtName[0] == 'x') &&
+        (ExtName.size() == 1 || isDigit(ExtName[1])))
       return createStringError(errc::invalid_argument,
-                               "'z' must be followed by a letter");
+                               "'" + Twine(ExtName[0]) +
+                                   "' must be followed by a letter");
 
     ISAInfo->addExtension(ExtName, {MajorVersion, MinorVersion});
   }

llvm/test/tools/llvm-objdump/ELF/RISCV/unknown-arch-attr.test

@@ -3,7 +3,7 @@
 ## The expected behavior is to ignore the unrecognized arch feature and
 ## continue to process the following arch features.
 ##
-## The object file has the "rv32i2p0_m2p0_x1p0" arch feature. "x1p0" is an
+## The object file has the "rv32i2p0_m2p0_y1p0" arch feature. "y1p0" is an
 ## unrecognized architecture extension. llvm-objdump will ignore it and decode
 ## "mul" instruction correctly according to "m2p0" in the arch feature.
 ##
@@ -34,5 +34,5 @@ Sections:
     Content: 3385C502
   - Name:    .riscv.attributes
     Type:    SHT_RISCV_ATTRIBUTES
-## The content is the encoding of the arch feature "rv32i2p0_m2p0_x1p0"
-    Content: 412300000072697363760001190000000572763332693270305F6D3270305F7831703000
+## The content is the encoding of the arch feature "rv32i2p0_m2p0_y1p0"
+    Content: 412300000072697363760001190000000572763332693270305F6D3270305F7931703000

llvm/unittests/TargetParser/RISCVISAInfoTest.cpp

@@ -62,6 +62,22 @@ TEST(ParseNormalizedArchString, RejectsBadZ) {
   }
 }
 
+TEST(ParseNormalizedArchString, RejectsBadS) {
+  for (StringRef Input : {"rv64i2p0_s1p0", "rv32i2p0_s2a1p0"}) {
+    EXPECT_EQ(
+        toString(RISCVISAInfo::parseNormalizedArchString(Input).takeError()),
+        "'s' must be followed by a letter");
+  }
+}
+
+TEST(ParseNormalizedArchString, RejectsBadX) {
+  for (StringRef Input : {"rv64i2p0_x1p0", "rv32i2p0_x2a1p0"}) {
+    EXPECT_EQ(
+        toString(RISCVISAInfo::parseNormalizedArchString(Input).takeError()),
+        "'x' must be followed by a letter");
+  }
+}
+
 TEST(ParseNormalizedArchString, AcceptsValidBaseISAsAndSetsXLen) {
   auto MaybeRV32I = RISCVISAInfo::parseNormalizedArchString("rv32i2p0");
   ASSERT_THAT_EXPECTED(MaybeRV32I, Succeeded());