q35: five seconds of boot spent waiting on disks that are not there

/proc/interrupts on q35 said "LOSING 526" ticks: 5 s of boot with
interrupts off. Timing each step of kmain() found two waits:

- AHCI sent a COMRESET to every empty port and waited a second for a
  link. DET = 0 is "no device", so only DET = 1 is worth a reset now
  (after 20 ms of grace for a drive that is still spinning up).
- ideinit() read 0xFF from a bus with no IDE controller, took the BSY
  bit for a busy drive, and waited out its 1 s timeout. A status of
  0xFF means no controller and is answered at once.

That is every PC with SATA disks. q35 now loses 33 ticks, under the
flag; i440fx still roots on IDE and loses none.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HH7xukt4aaiGj1nQEAoQf1

kernel/dev/ide.c

@@ -130,6 +130,18 @@ void
 ideinit(void)
 {
   initlock(&idelock, "ide");
+
+  // No controller at all: the bus floats and the status port reads 0xFF.
+  // That has BSY set, so idewait() below took it for a drive that was still
+  // busy and waited out its full second -- a second of every boot on q35 and
+  // on any PC whose disks are SATA, which is all of them. 0xFF is not a
+  // status any drive reports (BSY and DRDY are never both set), so it can be
+  // answered at once.
+  if(inb(0x1f7) == 0xFF){
+    kinfo("ide: no controller at the legacy ports\n");
+    return;
+  }
+
   // ISA signalling: edge-triggered, active high. Everything at a fixed
   // legacy port is wired this way; PCI devices are not.
   irq_register(IRQ_IDE, ide_irq, 0, "ide", 0);