From f0b99bc99a27c82d71db3e4f1de66fe1fd2c18fe Mon Sep 17 00:00:00 2001
From: DualFroz <me@dualfroz.com>
Date: Wed, 2 Sep 2026 19:23:37 +0200
Subject: [PATCH] feat: extend bmp with wordCountFor and bmpInit

---
 src/bmp.c     | 90 +++++++++++++++++++++++++++++++++++++++++++++++++++
 src/boot.S    | 20 ++++++++++++
 src/console.c | 16 ++++++---
 3 files changed, 122 insertions(+), 4 deletions(-)
 create mode 100644 src/bmp.c

diff --git a/src/bmp.c b/src/bmp.c
new file mode 100644
index 0000000..f5fd04e
--- /dev/null
+++ b/src/bmp.c
@@ -0,0 +1,90 @@
+#include "bmp.h"
+
+#define bmpBitsPerWord 64
+
+static usize wordCountFor(usize bitCount)
+{
+    return (bitCount + bmpBitsPerWord - 1) / bmpBitsPerWord;
+}
+
+void bmpInit(Bitmap* b, uint64* backingWords, usize bitCount)
+{
+    b->words = backingWords;
+    b->bitCount = bitCount;
+    bmpClearAll(b);
+}
+
+void bmpSet(Bitmap* b, usize index)
+{
+    if (index >= b->bitCount)
+        return;
+    b->words[index / bmpBitsPerWord] |= (1UL << (index % bmpBitsPerWord));
+}
+
+void bmpClear(Bitmap* b, usize index)
+{
+    if (index >= b->bitCount)
+        return;
+    b->words[index / bmpBitsPerWord] &= ~(1UL << (index % bmpBitsPerWord));
+}
+
+bool bmpTest(const Bitmap* b, usize index)
+{
+    if (index >= b->bitCount)
+        return false;
+    return (b->words[index / bmpBitsPerWord] & (1UL << (index % bmpBitsPerWord))) != 0;
+}
+
+void bmpSetRange(Bitmap* b, usize start, usize count)
+{
+    for (usize i = start; i < start + count && i < b->bitCount; i++)
+        bmpSet(b, i);
+}
+
+void bmpClearRange(Bitmap* b, usize start, usize count)
+{
+    for (usize i = start; i < start + count && i < b->bitCount; i++)
+        bmpClear(b, i);
+}
+
+void bmpSetAll(Bitmap* b)
+{
+    usize words = wordCountFor(b->bitCount);
+    for (usize i = 0; i < words; i++)
+        b->words[i] = ~0UL;
+}
+
+void bmpClearAll(Bitmap* b)
+{
+    usize words = wordCountFor(b->bitCount);
+    for (usize i = 0; i < words; i++)
+        b->words[i] = 0;
+}
+
+usize bmpCountSet(const Bitmap* b)
+{
+    usize total = 0;
+    for (usize i = 0; i < b->bitCount; i++) {
+        if (bmpTest(b, i))
+            total++;
+    }
+    return total;
+}
+
+isize bmpFindFirstClear(const Bitmap* b)
+{
+    for (usize i = 0; i < b->bitCount; i++) {
+        if (!bmpTest(b, i))
+            return (isize)i;
+    }
+    return -1;
+}
+
+isize bmpFindFirstSet(const Bitmap* b)
+{
+    for (usize i = 0; i < b->bitCount; i++) {
+        if (bmpTest(b, i))
+            return (isize)i;
+    }
+    return -1;
+}
diff --git a/src/boot.S b/src/boot.S
index a9ec92f..9474288 100644
--- a/src/boot.S
+++ b/src/boot.S
@@ -60,6 +60,10 @@ fillPd:
     movl $0xC0000080, %ecx
     rdmsr
     orl $(1 << 8), %eax
+    cmpl $0, cpuNxSupported
+    je 2f
+    orl $(1 << 11), %eax
+2:
     wrmsr
 
     movl %cr0, %eax
@@ -76,6 +80,16 @@ checkLongMode:
     jb noLongMode
     movl $0x80000001, %eax
     cpuid
+
+    /* Same cpuid leaf already gives us CPUID.80000001H:EDX.NX (bit 20) for
+       free -- stash it in cpuNxSupported (checked below, in %eax's later
+       rdmsr/wrmsr dance) before the LM (bit 29) test below, which does not
+       touch %edx, leaves it intact either way. */
+    movl $0, cpuNxSupported
+    testl $(1 << 20), %edx
+    jz 1f
+    movl $1, cpuNxSupported
+1:
     testl $(1 << 29), %edx
     jz noLongMode
     ret
@@ -111,6 +125,12 @@ bootGdtPointer:
     .word bootGdtEnd - bootGdtStart - 1
     .quad bootGdtStart
 
+.section .boot.data, "aw", @progbits
+.align 4
+.global cpuNxSupported
+cpuNxSupported:
+    .long 0
+
 .section .boot.pagetables, "aw", @progbits
 .align 4096
 .global bootPml4
diff --git a/src/console.c b/src/console.c
index 1d316cc..54dad47 100644
--- a/src/console.c
+++ b/src/console.c
@@ -43,13 +43,21 @@ static void writeUnsigned(uint64 value, uint32 base, bool upper, int minWidth, c
         kputchar(buffer[--length]);
 }
 
-static void writeSigned(int64 value)
+/* Mirrors klib.c's vsnprintf sinkPutSigned exactly (same minWidth/pad
+   contract, same "sign eats one column of the requested width" rule) --
+   found and fixed this fala after shell.c's new `ls -l` printed
+   "2026-9-7" instead of "2026-09-07": every prior kprintf call site that
+   wanted zero-padded fields (thread ids, sizes, hex addresses) happened
+   to use an unsigned specifier, so this signed-value width/pad bug went
+   completely unnoticed until the first zero-padded %d/%ld call ever
+   appeared. */
+static void writeSigned(int64 value, int minWidth, char pad)
 {
     if (value < 0) {
         kputchar('-');
-        writeUnsigned((uint64)(-value), 10, false, 0, ' ');
+        writeUnsigned((uint64)(-value), 10, false, minWidth > 0 ? minWidth - 1 : 0, pad);
     } else {
-        writeUnsigned((uint64)value, 10, false, 0, ' ');
+        writeUnsigned((uint64)value, 10, false, minWidth, pad);
     }
 }
 
@@ -93,7 +101,7 @@ void kprintf(const char* format, ...)
             break;
         case 'd':
         case 'i':
-            writeSigned(longCount ? va_arg(args, int64) : (int64)va_arg(args, int));
+            writeSigned(longCount ? va_arg(args, int64) : (int64)va_arg(args, int), width, pad);
             break;
         case 'u':
             writeUnsigned(longCount ? va_arg(args, uint64) : (uint64)va_arg(args, uint32), 10, false, width, pad);