[src/trunk]: src/sys/arch/i386/stand PR/51953: fix unable to boot on some AMD...

[nonaka <nonaka%NetBSD.org@localhost>]

details:   https://anonhg.NetBSD.org/src/rev/3ffb234566ea
branches:  trunk
changeset: 351376:3ffb234566ea
user:      nonaka <nonaka%NetBSD.org@localhost>
date:      Sat Feb 11 10:23:39 2017 +0000

description:
PR/51953: fix unable to boot on some AMD machine.

Delayed the timing to copy the kernel to actual address.
copy routine from common/lib/libc/arch/x86_64/string/bcopy.S

diffstat:

 sys/arch/i386/stand/efiboot/bootx64/efibootx64.c  |   11 +-
 sys/arch/i386/stand/efiboot/bootx64/startprog64.S |  123 ++++++++++++++++++++-
 sys/arch/i386/stand/efiboot/efiboot.c             |    4 +-
 sys/arch/i386/stand/efiboot/efiboot.h             |    4 +-
 sys/arch/i386/stand/lib/exec.c                    |    9 +-
 5 files changed, 132 insertions(+), 19 deletions(-)

diffs (249 lines):

diff -r a3d9e4a82f84 -r 3ffb234566ea sys/arch/i386/stand/efiboot/bootx64/efibootx64.c
--- a/sys/arch/i386/stand/efiboot/bootx64/efibootx64.c  Sat Feb 11 10:18:10 2017 +0000
+++ b/sys/arch/i386/stand/efiboot/bootx64/efibootx64.c  Sat Feb 11 10:23:39 2017 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: efibootx64.c,v 1.1 2017/01/24 11:09:14 nonaka Exp $    */
+/*     $NetBSD: efibootx64.c,v 1.2 2017/02/11 10:23:39 nonaka Exp $    */
 
 /*-
  * Copyright (c) 2016 Kimihiro Nonaka <nonaka%netbsd.org@localhost>
@@ -32,8 +32,10 @@
 
 struct x86_boot_params boot_params;
 
-void startprog64_start(void *, physaddr_t, physaddr_t);
-extern void (*startprog64)(void *, physaddr_t, physaddr_t);
+void startprog64_start(physaddr_t, physaddr_t, physaddr_t, u_long,
+    void *, physaddr_t);
+extern void (*startprog64)(physaddr_t, physaddr_t, physaddr_t, u_long,
+    void *, physaddr_t);
 extern u_int startprog64_size;
 
 void
@@ -64,7 +66,8 @@
                memcpy(newsp, argv, sizeof(*argv) * argc);
        }
 
-       (*startprog64)(startprog64, entry, (physaddr_t)newsp);
+       (*startprog64)(efi_kernel_start, efi_kernel_start + efi_loadaddr,
+           (physaddr_t)newsp, efi_kernel_size, startprog64, entry);
 }
 
 /* ARGSUSED */
diff -r a3d9e4a82f84 -r 3ffb234566ea sys/arch/i386/stand/efiboot/bootx64/startprog64.S
--- a/sys/arch/i386/stand/efiboot/bootx64/startprog64.S Sat Feb 11 10:18:10 2017 +0000
+++ b/sys/arch/i386/stand/efiboot/bootx64/startprog64.S Sat Feb 11 10:23:39 2017 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: startprog64.S,v 1.2 2017/02/04 16:14:27 christos Exp $ */
+/*     $NetBSD: startprog64.S,v 1.3 2017/02/11 10:23:39 nonaka Exp $   */
 /*     NetBSD: startprog.S,v 1.3 2003/02/01 14:48:18 dsl Exp   */
 
 /* starts program in protected mode / flat space
@@ -78,26 +78,135 @@
        .p2align 4,,15
 
 /*
- * startprog64(loadddr,entry,stack)
+ * startprog64(loadddr,entry,stack,kern_load,kern_start,kern_size)
  */
 ENTRY(startprog64_start)
 start:
        /*
         * This function is to call the loaded kernel's start() with
         * 32bit segment mode from x64 mode.
-        * %rdi: loaded start address
-        * %rsi: kernel entry address
+        * %rdi: kernel start address
+        * %rsi: loaded kernel address
         * %rdx: stack address
+        * %rcx: loaded kernel size
+        * %r8 : loaded start address
+        * %r9 : kernel entry address
         */
 
-       cld             # LynxOS depends on it
+       cld             /* LynxOS depends on it */
+
+       cli
+
+       /* Copy kernel */
+       mov     %rcx, %r12              /* original kernel size */
+       movq    %rdi, %r11              /* for misaligned check */
+
+#if !defined(NO_OVERLAP)
+       movq    %rdi, %r13
+       subq    %rsi, %r13
+#endif
+
+       shrq    $3, %rcx                /* count for copy by words */
+       jz      8f                      /* j if less than 8 bytes */
+
+       lea     -8(%rdi, %r12), %r14    /* target address of last 8 */
+       mov     -8(%rsi, %r12), %r15    /* get last word */
+#if !defined(NO_OVERLAP)
+       cmpq    %r12, %r13              /* overlapping? */
+       jb      10f
+#endif
+
+/*
+ * Non-overlaping, copy forwards.
+ * Newer Intel cpus (Nehalem) will do 16byte read/write transfers
+ * if %ecx is more than 76.
+ * AMD might do something similar some day.
+ */
+       and     $7, %r11                /* destination misaligned ? */
+       jnz     2f
+       rep
+       movsq
+       mov     %r15, (%r14)            /* write last word */
+       jmp     .Lcopy_done
+
+/*
+ * Destination misaligned
+ * AMD say it is better to align the destination (not the source).
+ * This will also re-align copies if the source and dest are both
+ * misaligned by the same amount)
+ * (I think Nehalem will use its accelerated copy if the source
+ * and destination have the same alignment.)
+ */
+2:
+       lea     -9(%r11, %r12), %rcx    /* post re-alignment count */
+       neg     %r11                    /* now -1 .. -7 */
+       mov     (%rsi), %r12            /* get first word */
+       mov     %rdi, %r13              /* target for first word */
+       lea     8(%rsi, %r11), %rsi
+       lea     8(%rdi, %r11), %rdi
+       shr     $3, %rcx
+       rep
+       movsq
+       mov     %r12, (%r13)            /* write first word */
+       mov     %r15, (%r14)            /* write last word */
+       jmp     .Lcopy_done
+
+#if !defined(NO_OVERLAP)
+/* Must copy backwards.
+ * Reverse copy is probably easy to code faster than 'rep movds'
+ * since that requires (IIRC) an extra clock every 3 iterations (AMD).
+ * However I don't suppose anything cares that much!
+ * The big cost is the std/cld pair - reputedly 50+ cycles on Netburst P4.
+ * The copy is aligned with the buffer start (more likely to
+ * be a multiple of 8 than the end).
+ */
+10:
+       lea     -8(%rsi, %rcx, 8), %rsi
+       lea     -8(%rdi, %rcx, 8), %rdi
+       std
+       rep
+       movsq
+       cld
+       mov     %r15, (%r14)    /* write last bytes */
+       jmp     .Lcopy_done
+#endif
+
+/* Less than 8 bytes to copy, copy by bytes */
+/* Intel Nehalem optimise 'rep movsb' for <= 7 bytes (9-15 clocks).
+ * For longer transfers it is 50+ !
+ */
+8:     mov     %r12, %rcx
+
+#if !defined(NO_OVERLAP)
+       cmpq    %r12, %r13      /* overlapping? */
+       jb      81f
+#endif
+
+       /* nope, copy forwards. */
+       rep
+       movsb
+       jmp     .Lcopy_done
+
+#if !defined(NO_OVERLAP)
+/* Must copy backwards */
+81:
+       lea     -1(%rsi, %rcx), %rsi
+       lea     -1(%rdi, %rcx), %rdi
+       std
+       rep
+       movsb
+       cld
+#endif
+       /* End of copy kernel */
+.Lcopy_done:
+
+       mov     %r8, %rdi       /* %rdi: loaded start address */
+       mov     %r9, %rsi       /* %rsi: kernel entry address */
 
        /* Prepare jump address */
        lea     (start32a - start)(%rdi), %rax
        movl    %eax, (start32r - start)(%rdi)
 
-       cli
-
        /* Setup GDT */
        lea     (gdt - start)(%rdi), %rax
        mov     %rax, (gdtrr - start)(%rdi)
diff -r a3d9e4a82f84 -r 3ffb234566ea sys/arch/i386/stand/efiboot/efiboot.c
--- a/sys/arch/i386/stand/efiboot/efiboot.c     Sat Feb 11 10:18:10 2017 +0000
+++ b/sys/arch/i386/stand/efiboot/efiboot.c     Sat Feb 11 10:23:39 2017 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: efiboot.c,v 1.3 2017/02/11 10:15:55 nonaka Exp $       */
+/*     $NetBSD: efiboot.c,v 1.4 2017/02/11 10:23:39 nonaka Exp $       */
 
 /*-
  * Copyright (c) 2016 Kimihiro Nonaka <nonaka%netbsd.org@localhost>
@@ -35,6 +35,8 @@
 EFI_DEVICE_PATH *efi_bootdp;
 EFI_LOADED_IMAGE *efi_li;
 uintptr_t efi_main_sp;
+physaddr_t efi_loadaddr, efi_kernel_start;
+u_long efi_kernel_size;
 bool efi_cleanuped;
 
 static EFI_PHYSICAL_ADDRESS heap_start = EFI_ALLOCATE_MAX_ADDRESS;
diff -r a3d9e4a82f84 -r 3ffb234566ea sys/arch/i386/stand/efiboot/efiboot.h
--- a/sys/arch/i386/stand/efiboot/efiboot.h     Sat Feb 11 10:18:10 2017 +0000
+++ b/sys/arch/i386/stand/efiboot/efiboot.h     Sat Feb 11 10:23:39 2017 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: efiboot.h,v 1.3 2017/02/11 10:15:55 nonaka Exp $       */
+/*     $NetBSD: efiboot.h,v 1.4 2017/02/11 10:23:39 nonaka Exp $       */
 
 /*-
  * Copyright (c) 2016 Kimihiro Nonaka <nonaka%netbsd.org@localhost>
@@ -48,6 +48,8 @@
 extern EFI_DEVICE_PATH *efi_bootdp;
 extern EFI_LOADED_IMAGE *efi_li;
 extern uintptr_t efi_main_sp;
+extern physaddr_t efi_loadaddr, efi_kernel_start;
+extern u_long efi_kernel_size;
 extern bool efi_cleanuped;
 void efi_cleanup(void);
 
diff -r a3d9e4a82f84 -r 3ffb234566ea sys/arch/i386/stand/lib/exec.c
--- a/sys/arch/i386/stand/lib/exec.c    Sat Feb 11 10:18:10 2017 +0000
+++ b/sys/arch/i386/stand/lib/exec.c    Sat Feb 11 10:23:39 2017 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: exec.c,v 1.64 2017/02/11 10:18:10 nonaka Exp $  */
+/*     $NetBSD: exec.c,v 1.65 2017/02/11 10:23:39 nonaka Exp $  */
 
 /*
  * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc.
@@ -112,7 +112,6 @@
 #ifdef EFIBOOT
 #include "efiboot.h"
 #undef DEBUG   /* XXX */
-static u_long efi_loadaddr;
 #endif
 
 #define BOOT_NARGS     6
@@ -460,10 +459,8 @@
                bootinfo->entry[i] = vtophys(p);
        }
 
-       /* Copy the kernel to original load address. */
-       memmove((void *)marks[MARK_START],
-           (void *)(efi_loadaddr + marks[MARK_START]),
-           marks[MARK_END] - marks[MARK_START]);
+       efi_kernel_start = marks[MARK_START];
+       efi_kernel_size = marks[MARK_END] - efi_kernel_start;
 #endif
        startprog(marks[MARK_ENTRY], BOOT_NARGS, boot_argv,
            x86_trunc_page(basemem * 1024));