Merge branch 'master' into master

api/docs/code_reviews.dox

@@ -198,10 +198,11 @@ appear.  Do not simply reply at the top level, as such comments are more
 difficult to follow as a conversation thread with context.
 
 After making changes in response to review comments, commit those changes
-locally as a new commit on top of your original commit.  If the reviewer
-requested changes to the commit message, use a new proposed final commit
-message as the message for this change commit so that the reviewer can
-review the new message.
+locally as a new commit on top of your original commit (do not squash the
+new changes into the original commit; use separate commits).  If the reviewer
+requested changes to the commit message or title, edit those directly in
+the pull request text boxes as those are what become the final commit
+message.
 
 Project members should then push the new commit to the pull request via:
 
@@ -215,7 +216,10 @@ separate commit (these separate commits will all be squashed together upon mergi
 When the requested changes have been pushed, request a re-review from the reviewer so they know that the pull request is ready for another round of reviewing.
 This can be done by clicking the re-review button next to the reviewer's name at the top of the right sidebar on the pull request page.
 
-The final squash-and-merge step will squash these additional commits with the original to make a single commit that will be fast-forward-merged into master (see below for more details).
+The final squash-and-merge step will squash these additional commits with the
+original to make a single commit that will be fast-forward-merged into master (see
+below for more details).  The final master branch merge commit's title and
+description come directly from the pull request title and description.
 
 # Updating from Master
 

core/ir/x86/decode_table.c

@@ -6159,20 +6159,20 @@ const instr_info_t e_vex_extensions[][3] = {
     {INVALID, 0x6638db18, catUncategorized, "(bad)", xx, xx, xx, xx, xx, no, x, NA},
   }, { /* e_vex ext 32 */
     {OP_aesenc,  0x6638dc18, catSIMD, "aesenc",  Vdq, xx, Wdq,Vdq, xx, mrm|reqp, x, END_LIST},
-    {OP_vaesenc, 0x6638dc18, catSIMD, "vaesenc",  Vx, xx, Hx,Wx, xx, mrm|vex|reqp, x, END_LIST},
-    {INVALID, 0x6638dc18, catUncategorized, "(bad)", xx, xx, xx, xx, xx, no, x, NA},
+    {OP_vaesenc, 0x6638dc18, catSIMD, "vaesenc",  Vx, xx, Hx,Wx, xx, mrm|vex|reqp, x, tvex[32][2]},
+    {OP_vaesenc, 0x6638dc18, catSIMD, "vaesenc",  Ve, xx, He,We, xx, mrm|evex|reqp|ttfvm, x, END_LIST},
   }, { /* e_vex ext 33 */
     {OP_aesenclast,  0x6638dd18, catSIMD, "aesenclast",Vdq,xx,Wdq,Vdq,xx, mrm|reqp, x, END_LIST},
-    {OP_vaesenclast, 0x6638dd18, catSIMD, "vaesenclast",Vx,xx,Hx,Wx,xx, mrm|vex|reqp, x, END_LIST},
-    {INVALID, 0x6638dd18, catUncategorized, "(bad)", xx, xx, xx, xx, xx, no, x, NA},
+    {OP_vaesenclast, 0x6638dd18, catSIMD, "vaesenclast",Vx,xx,Hx,Wx,xx, mrm|vex|reqp, x, tvex[33][2]},
+    {OP_vaesenclast, 0x6638dd18, catSIMD, "vaesenclast",Ve,xx,He,We,xx, mrm|evex|reqp|ttfvm, x, END_LIST},
   }, { /* e_vex ext 34 */
     {OP_aesdec,  0x6638de18, catSIMD, "aesdec",  Vdq, xx, Wdq,Vdq, xx, mrm|reqp, x, END_LIST},
-    {OP_vaesdec, 0x6638de18, catSIMD, "vaesdec",  Vx, xx, Hx,Wx, xx, mrm|vex|reqp, x, END_LIST},
-    {INVALID, 0x6638de18, catUncategorized, "(bad)", xx, xx, xx, xx, xx, no, x, NA},
+    {OP_vaesdec, 0x6638de18, catSIMD, "vaesdec",  Vx, xx, Hx,Wx, xx, mrm|vex|reqp, x, tvex[34][2]},
+    {OP_vaesdec, 0x6638de18, catSIMD, "vaesdec",  Ve, xx, He,We, xx, mrm|evex|reqp|ttfvm, x, END_LIST},
   }, { /* e_vex ext 35 */
     {OP_aesdeclast,  0x6638df18, catSIMD, "aesdeclast",Vdq,xx,Wdq,Vdq,xx, mrm|reqp, x, END_LIST},
-    {OP_vaesdeclast, 0x6638df18, catSIMD, "vaesdeclast",Vx,xx,Hx,Wx,xx, mrm|vex|reqp, x, END_LIST},
-    {INVALID, 0x6638df18, catUncategorized, "(bad)", xx, xx, xx, xx, xx, no, x, NA},
+    {OP_vaesdeclast, 0x6638df18, catSIMD, "vaesdeclast",Vx,xx,Hx,Wx,xx, mrm|vex|reqp, x, tvex[35][2]},
+    {OP_vaesdeclast, 0x6638df18, catSIMD, "vaesdeclast",Ve,xx,He,We,xx, mrm|evex|reqp|ttfvm, x, END_LIST},
   }, { /* e_vex ext 36 */
     {OP_pextrb,   0x663a1418, catSIMD, "pextrb", Rd_Mb, xx, Vb_dq, Ib, xx, mrm|reqp, x, END_LIST},
     {OP_vpextrb,  0x663a1418, catSIMD, "vpextrb", Rd_Mb, xx, Vb_dq, Ib, xx, mrm|vex|reqp, x, tvex[36][2]},
@@ -6260,8 +6260,8 @@ const instr_info_t e_vex_extensions[][3] = {
     {INVALID, 0x663a6318, catUncategorized, "(bad)", xx, xx, xx, xx, xx, no, x, NA},
   }, { /* e_vex ext 57 */
     {OP_pclmulqdq, 0x663a4418, catSIMD, "pclmulqdq", Vdq, xx, Wdq, Ib, Vdq, mrm|reqp, x, END_LIST},
-    {OP_vpclmulqdq,0x663a4418, catSIMD, "vpclmulqdq", Vx, xx, Hx, Wx, Ib, mrm|vex|reqp, x, END_LIST},
-    {INVALID, 0x663a4418, catUncategorized, "(bad)", xx, xx, xx, xx, xx, no, x, NA},
+    {OP_vpclmulqdq,0x663a4418, catSIMD, "vpclmulqdq", Vx, xx, Hx, Wx, Ib, mrm|vex|reqp, x, tvex[57][2]},
+    {OP_vpclmulqdq,0x663a4418, catSIMD, "vpclmulqdq", Ve, xx, He, We, Ib, mrm|evex|reqp|ttfvm, x, END_LIST},
   }, { /* e_vex ext 58 */
     {OP_aeskeygenassist, 0x663adf18, catSIMD, "aeskeygenassist",Vdq,xx,Wdq,Ib,xx,mrm|reqp,x,END_LIST},
     {OP_vaeskeygenassist,0x663adf18, catSIMD, "vaeskeygenassist",Vdq,xx,Wdq,Ib,xx,mrm|vex|reqp,x,END_LIST},

core/unix/coredump.c

@@ -51,6 +51,7 @@
 #    error Unsupported architecture
 #endif
 
+#define MAX_BUFFER_SIZE 4096
 #define MAX_SECTION_HEADERS 300
 #define MAX_SECTION_NAME_BUFFER_SIZE 8192
 #define SECTION_HEADER_TABLE ".shstrtab"
@@ -484,11 +485,23 @@ os_dump_core_internal(dcontext_t *dcontext)
     // of LOAD program header (number of sections minus one since shstrtab does not
     // require a LOAD header).
     const ELF_OFF program_header_count = section_count;
+    // core_file_offset is the current offset of the core file to write the next
+    // section data.
+    ELF_OFF core_file_offset = sizeof(ELF_HEADER_TYPE) +
+        sizeof(ELF_PROGRAM_HEADER_TYPE) * program_header_count +
+        PROGRAM_HEADER_NOTE_LENGTH;
+    // We add padding to the core file so that loadable segments in the
+    // core file are page size aligned. The p_align field in the program header
+    // is applied to both memory and the offset to the core file.
+    const size_t loadable_segment_padding =
+        ALIGN_FORWARD(core_file_offset, os_page_size()) - core_file_offset;
+    core_file_offset = ALIGN_FORWARD(core_file_offset, os_page_size());
+
     if (!write_elf_header(elf_file, /*entry_point=*/(uint64_t)mc.pc,
-                          /*section_header_table_offset*/ sizeof(ELF_HEADER_TYPE) +
+                          /*section_header_table_offset=*/sizeof(ELF_HEADER_TYPE) +
                               PROGRAM_HEADER_NOTE_LENGTH +
                               sizeof(ELF_PROGRAM_HEADER_TYPE) * program_header_count +
-                              section_data_size,
+                              loadable_segment_padding + section_data_size,
                           /*flags=*/0,
                           /*program_header_count=*/program_header_count,
                           /*section_header_count=*/section_count,
@@ -509,9 +522,6 @@ os_dump_core_internal(dcontext_t *dcontext)
         return false;
     }
     // Write loadable program segment program headers.
-    ELF_OFF file_offset = sizeof(ELF_HEADER_TYPE) +
-        sizeof(ELF_PROGRAM_HEADER_TYPE) * program_header_count +
-        PROGRAM_HEADER_NOTE_LENGTH;
     // TODO i#7046: Merge adjacent sections with the same prot values.
     // The last section is shstrtab which stores the section names and it does
     // not require a LOAD program header.
@@ -530,7 +540,7 @@ os_dump_core_internal(dcontext_t *dcontext)
             section_header_info[section_index].vm_start;
         if (!write_program_header(
                 elf_file, PT_LOAD, flags,
-                /*offset=*/file_offset,
+                /*offset=*/core_file_offset,
                 /*virtual_address=*/(ELF_ADDR)section_header_info[section_index].vm_start,
                 /*physical_address=*/
                 (ELF_ADDR)section_header_info[section_index].vm_start,
@@ -540,7 +550,7 @@ os_dump_core_internal(dcontext_t *dcontext)
             os_close(elf_file);
             return false;
         }
-        file_offset += section_header_info[section_index].vm_end -
+        core_file_offset += section_header_info[section_index].vm_end -
             section_header_info[section_index].vm_start;
     }
     if (!write_prstatus_note(&mc, elf_file)) {
@@ -551,6 +561,20 @@ os_dump_core_internal(dcontext_t *dcontext)
         os_close(elf_file);
         return false;
     }
+    // Add padding to the core file such that loadable segments are aligned to
+    // the page size in the core file.
+    char buffer[MAX_BUFFER_SIZE];
+    size_t remaining_bytes = loadable_segment_padding;
+    memset(buffer, 0, MIN(MAX_BUFFER_SIZE, loadable_segment_padding));
+    while (remaining_bytes > 0) {
+        const size_t bytes = MIN(remaining_bytes, MAX_BUFFER_SIZE);
+        if (os_write(elf_file, buffer, bytes) != bytes) {
+            SYSLOG_INTERNAL_ERROR("Failed to add padding to the core dump file.");
+            os_close(elf_file);
+            return false;
+        }
+        remaining_bytes -= bytes;
+    }
     // Write memory content to the core dump file.
     for (int section_index = 0; section_index < section_count - 1; ++section_index) {
         const size_t length = section_header_info[section_index].vm_end -
@@ -577,9 +601,9 @@ os_dump_core_internal(dcontext_t *dcontext)
         return false;
     }
     // Write section headers to the core dump file.
-    file_offset = sizeof(ELF_HEADER_TYPE) +
+    core_file_offset = sizeof(ELF_HEADER_TYPE) +
         sizeof(ELF_PROGRAM_HEADER_TYPE) * program_header_count +
-        PROGRAM_HEADER_NOTE_LENGTH;
+        PROGRAM_HEADER_NOTE_LENGTH + loadable_segment_padding;
 
     // The section_count includes the section name section, so we need to skip
     // it in the loop. The section name section is handled differently after
@@ -594,7 +618,8 @@ os_dump_core_internal(dcontext_t *dcontext)
         }
         if (!write_section_header(
                 elf_file, section_header_info[section_index].name_offset, SHT_PROGBITS,
-                flags, (ELF_ADDR)section_header_info[section_index].vm_start, file_offset,
+                flags, (ELF_ADDR)section_header_info[section_index].vm_start,
+                core_file_offset,
                 section_header_info[section_index].vm_end -
                     section_header_info[section_index].vm_start,
                 /*link=*/0,
@@ -603,13 +628,13 @@ os_dump_core_internal(dcontext_t *dcontext)
             os_close(elf_file);
             return false;
         }
-        file_offset += section_header_info[section_index].vm_end -
+        core_file_offset += section_header_info[section_index].vm_end -
             section_header_info[section_index].vm_start;
     }
     // Write the section name section.
     if (!write_section_header(
             elf_file, string_table_offset - strlen(SECTION_HEADER_TABLE), SHT_STRTAB,
-            /*flags=*/0, /*virtual_address=*/0, file_offset,
+            /*flags=*/0, /*virtual_address=*/0, core_file_offset,
             /*section_size=*/string_table_offset, /*link=*/0,
             /*info=*/0, /*alignment=*/1,
             /*entry_size=*/0)) {

suite/tests/api/ir_x86_3args_avx512_evex.h

@@ -215,3 +215,67 @@ OPCODE(vpsadbw_zhizhizhi, vpsadbw, vpsadbw, X64_ONLY, REGARG(ZMM16), REGARG(ZMM1
        REGARG(ZMM31))
 OPCODE(vpsadbw_zhizhild, vpsadbw, vpsadbw, X64_ONLY, REGARG(ZMM16), REGARG(ZMM31),
        MEMARG(OPSZ_64))
+OPCODE(vaesenc_evex_xhixhixhi, vaesenc, vaesenc, X64_ONLY, REGARG(XMM16), REGARG(XMM17),
+       REGARG(XMM18))
+OPCODE(vaesenc_evex_xhixhild, vaesenc, vaesenc, X64_ONLY, REGARG(XMM16), REGARG(XMM17),
+       MEMARG(OPSZ_16))
+OPCODE(vaesenc_evex_yhiyhiyhi, vaesenc, vaesenc, X64_ONLY, REGARG(YMM16), REGARG(YMM17),
+       REGARG(YMM18))
+OPCODE(vaesenc_evex_yhiyhild, vaesenc, vaesenc, X64_ONLY, REGARG(YMM16), REGARG(YMM17),
+       MEMARG(OPSZ_32))
+OPCODE(vaesenc_evex_zlozlozlo, vaesenc, vaesenc, 0, REGARG(ZMM0), REGARG(ZMM1),
+       REGARG(ZMM2))
+OPCODE(vaesenc_evex_zlozlold, vaesenc, vaesenc, 0, REGARG(ZMM0), REGARG(ZMM1),
+       MEMARG(OPSZ_64))
+OPCODE(vaesenc_evex_zhizhizhi, vaesenc, vaesenc, X64_ONLY, REGARG(ZMM16), REGARG(ZMM17),
+       REGARG(ZMM18))
+OPCODE(vaesenc_evex_zhizhild, vaesenc, vaesenc, X64_ONLY, REGARG(ZMM16), REGARG(ZMM17),
+       MEMARG(OPSZ_64))
+OPCODE(vaesenclast_evex_xhixhixhi, vaesenclast, vaesenclast, X64_ONLY, REGARG(XMM16),
+       REGARG(XMM17), REGARG(XMM18))
+OPCODE(vaesenclast_evex_xhixhild, vaesenclast, vaesenclast, X64_ONLY, REGARG(XMM16),
+       REGARG(XMM17), MEMARG(OPSZ_16))
+OPCODE(vaesenclast_evex_yhiyhiyhi, vaesenclast, vaesenclast, X64_ONLY, REGARG(YMM16),
+       REGARG(YMM17), REGARG(YMM15))
+OPCODE(vaesenclast_evex_yhiyhild, vaesenclast, vaesenclast, X64_ONLY, REGARG(YMM16),
+       REGARG(YMM17), MEMARG(OPSZ_32))
+OPCODE(vaesenclast_evex_zlozlozlo, vaesenclast, vaesenclast, 0, REGARG(ZMM0),
+       REGARG(ZMM1), REGARG(ZMM2))
+OPCODE(vaesenclast_evex_zlozlold, vaesenclast, vaesenclast, 0, REGARG(ZMM0), REGARG(ZMM1),
+       MEMARG(OPSZ_64))
+OPCODE(vaesenclast_evex_zhizhizhi, vaesenclast, vaesenclast, X64_ONLY, REGARG(ZMM16),
+       REGARG(ZMM17), REGARG(ZMM15))
+OPCODE(vaesenclast_evex_zhizhild, vaesenclast, vaesenclast, X64_ONLY, REGARG(ZMM16),
+       REGARG(ZMM17), MEMARG(OPSZ_64))
+OPCODE(vaesdec_evex_xhixhixlo, vaesdec, vaesdec, X64_ONLY, REGARG(XMM16), REGARG(XMM17),
+       REGARG(XMM0))
+OPCODE(vaesdec_evex_xhixhild, vaesdec, vaesdec, X64_ONLY, REGARG(XMM16), REGARG(XMM17),
+       MEMARG(OPSZ_16))
+OPCODE(vaesdec_evex_yhiyhiylo, vaesdec, vaesdec, X64_ONLY, REGARG(YMM16), REGARG(YMM17),
+       REGARG(YMM0))
+OPCODE(vaesdec_evex_yhiyhild, vaesdec, vaesdec, X64_ONLY, REGARG(YMM16), REGARG(YMM17),
+       MEMARG(OPSZ_32))
+OPCODE(vaesdec_evex_zlozlozlo, vaesdec, vaesdec, 0, REGARG(ZMM0), REGARG(ZMM1),
+       REGARG(ZMM2))
+OPCODE(vaesdec_evex_zlozlold, vaesdec, vaesdec, 0, REGARG(ZMM0), REGARG(ZMM1),
+       MEMARG(OPSZ_64))
+OPCODE(vaesdec_evex_zhizhizlo, vaesdec, vaesdec, X64_ONLY, REGARG(ZMM16), REGARG(ZMM17),
+       REGARG(ZMM15))
+OPCODE(vaesdec_evex_zhizhild, vaesdec, vaesdec, X64_ONLY, REGARG(ZMM8), REGARG(ZMM9),
+       MEMARG(OPSZ_64))
+OPCODE(vaesdeclast_evex_xhixhixhi, vaesdeclast, vaesdeclast, X64_ONLY, REGARG(XMM16),
+       REGARG(XMM17), REGARG(XMM18))
+OPCODE(vaesdeclast_evex_xhixhild, vaesdeclast, vaesdeclast, X64_ONLY, REGARG(XMM16),
+       REGARG(XMM17), MEMARG(OPSZ_16))
+OPCODE(vaesdeclast_evex_yhiyhiyhi, vaesdeclast, vaesdeclast, X64_ONLY, REGARG(YMM16),
+       REGARG(YMM17), REGARG(YMM31))
+OPCODE(vaesdeclast_evex_yhiyhild, vaesdeclast, vaesdeclast, X64_ONLY, REGARG(YMM16),
+       REGARG(YMM17), MEMARG(OPSZ_32))
+OPCODE(vaesdeclast_evex_zlozlozlo, vaesdeclast, vaesdeclast, 0, REGARG(ZMM0),
+       REGARG(ZMM1), REGARG(ZMM2))
+OPCODE(vaesdeclast_evex_zlozlold, vaesdeclast, vaesdeclast, 0, REGARG(ZMM0), REGARG(ZMM1),
+       MEMARG(OPSZ_64))
+OPCODE(vaesdeclast_evex_zhizhizhi, vaesdeclast, vaesdeclast, X64_ONLY, REGARG(ZMM16),
+       REGARG(ZMM17), REGARG(ZMM31))
+OPCODE(vaesdeclast_evex_zhizhild, vaesdeclast, vaesdeclast, X64_ONLY, REGARG(ZMM16),
+       REGARG(ZMM17), MEMARG(OPSZ_64))

suite/tests/api/ir_x86_4args_avx512_evex.h

@@ -67,3 +67,17 @@ OPCODE(vpinsrq_xhixhir, vpinsrq, vpinsrq, X64_ONLY, REGARG(XMM16),
        REGARG_PARTIAL(XMM31, OPSZ_8), REGARG(RAX), IMMARG(OPSZ_1))
 OPCODE(vpinsrq_xhixhim, vpinsrq, vpinsrq, X64_ONLY, REGARG(XMM16),
        REGARG_PARTIAL(XMM31, OPSZ_8), MEMARG(OPSZ_8), IMMARG(OPSZ_1))
+OPCODE(vpclmulqdq_evex_xhixhixhi, vpclmulqdq, vpclmulqdq, X64_ONLY, REGARG(XMM16),
+       REGARG(XMM17), REGARG(XMM18), IMMARG(OPSZ_1))
+OPCODE(vpclmulqdq_evex_xhixhild, vpclmulqdq, vpclmulqdq, X64_ONLY, REGARG(XMM16),
+       REGARG(XMM17), MEMARG(OPSZ_16), IMMARG(OPSZ_1))
+OPCODE(vpclmulqdq_evex_yhiyhiyhi, vpclmulqdq, vpclmulqdq, X64_ONLY, REGARG(YMM16),
+       REGARG(YMM17), REGARG(YMM18), IMMARG(OPSZ_1))
+OPCODE(vpclmulqdq_evex_yhiyhild, vpclmulqdq, vpclmulqdq, X64_ONLY, REGARG(YMM16),
+       REGARG(YMM17), MEMARG(OPSZ_32), IMMARG(OPSZ_1))
+OPCODE(vpclmulqdq_zlozlozlo, vpclmulqdq, vpclmulqdq, 0, REGARG(ZMM0), REGARG(ZMM1),
+       REGARG(ZMM2), IMMARG(OPSZ_1))
+OPCODE(vpclmulqdq_zlozlold, vpclmulqdq, vpclmulqdq, 0, REGARG(ZMM0), REGARG(ZMM1),
+       MEMARG(OPSZ_64), IMMARG(OPSZ_1))
+OPCODE(vpclmulqdq_zhizhizhi, vpclmulqdq, vpclmulqdq, X64_ONLY, REGARG(ZMM16),
+       REGARG(ZMM17), REGARG(ZMM18), IMMARG(OPSZ_1))