refactor: ELF step uses FSBL instead of ps7_init

Problem: dow app.elf to DDR (0x100000) failed with MMU section
translation fault because the CPU had MMU enabled from previously
executed Flash code.  ps7_init could reset this but user wants
FSBL-based flow.

Fix: run_elf() now downloads FSBL first (to OCM 0x0 — always
accessible), runs it to initialize clocks/DDR/MIO and force JTAG
boot mode, then downloads the app ELF to DDR.

New flow:
  connect → targets 2 → dow fsbl.elf → con → after 3s → stop
  → dow app.elf → mwr 0xF800025C 0x1 → con

Fallback: if no FSBL configured, use direct dow (legacy path).

Removed: _run_elf_with_xsct (2-attempt ps7_init recovery) replaced
by _run_elf_via_fsbl and _run_elf_direct.

src/bitstream_programmer.py

@@ -329,7 +329,9 @@ def run_elf(
 ) -> BitstreamResult:
     """Download and run a .elf executable on the Zynq PS ARM core.
 
-    Uses xsct's `dow` (download) command, with Vivado as fallback.
+    First downloads FSBL (First Stage Bootloader) to OCM (0x0) to
+    initialize clocks, DDR, MIO and force JTAG boot mode.  Then
+    downloads the target ELF to DDR.
 
     Args:
         config: Application configuration.
@@ -347,10 +349,15 @@ def run_elf(
             message=f"ELF file not found: {elf_path}",
         )
 
+    fsbl_path = config.fsbl_elf_path if hasattr(config, 'fsbl_elf_path') else ""
+    if fsbl_path and not Path(fsbl_path).exists():
+        if callback:
+            callback("warning", f"FSBL not found: {fsbl_path}, will try direct dow")
+        fsbl_path = ""
+
     if callback:
         callback("start", f"Running ELF: {elf_path.name}...")
 
-    # Use xsct for CPU operations (Vivado HM cannot program PS CPU)
     xsct = get_xsct_path(
         vitis_path=config.vitis_path if hasattr(config, 'vitis_path') else "",
         xilinx_root=config.xilinx_path if hasattr(config, 'xilinx_path') else "",
@@ -362,30 +369,32 @@ def run_elf(
             message="xsct not found in PATH (required for CPU operations)",
         )
 
-    # Auto-detect ps7_init.tcl from config or BIT file directory
-    ps7_tcl = _find_ps7_init_tcl(config)
     timeout = config.step_timeouts.get("elf_download", 60)
-    return _run_elf_with_xsct(xsct, elf_path, callback, ps7_tcl, timeout)
+
+    if fsbl_path:
+        return _run_elf_via_fsbl(xsct, elf_path, fsbl_path, callback, timeout)
+    else:
+        return _run_elf_direct(xsct, elf_path, callback, timeout)
 
 
-def _run_elf_with_xsct(
+def _run_elf_via_fsbl(
     xsct_path: str,
     elf_path: Path,
+    fsbl_path: str,
     callback: ProgressCallback = None,
-    ps7_init_tcl: str = "",
     timeout: int = 60,
 ) -> BitstreamResult:
-    """Download and run ELF using xsct's dow command.
+    """Download and run ELF via FSBL bootloader.
 
-    PS was already initialized by the prior BIT download step, so the
-    DAP should be alive. First attempt skips ps7_init to avoid resetting
-    PS state. If DAP is broken, falls back to re-init PS.
+    FSBL loads to OCM (0x0) which is always accessible even when
+    MMU is enabled.  FSBL initializes clocks/DDR/MIO and forces
+    JTAG boot mode, making DDR accessible for the app ELF.
 
     Args:
         xsct_path: Path to xsct executable.
-        elf_path: Path to .elf file.
+        elf_path: Path to target .elf file.
+        fsbl_path: Path to FSBL .elf file.
         callback: Optional progress callback.
-        ps7_init_tcl: Optional ps7_init.tcl path for recovery.
         timeout: Subprocess timeout in seconds.
 
     Returns:
@@ -393,99 +402,132 @@ def _run_elf_with_xsct(
     """
     import tempfile, os
 
-    has_ps7 = bool(ps7_init_tcl and Path(ps7_init_tcl).exists())
-
-    def _run(tcl: str) -> tuple[bool, str]:
-        """Run TCL via xsct, return (success, output)."""
-        with tempfile.NamedTemporaryFile(
-            mode="w", suffix=".tcl", delete=False
-        ) as tf:
-            tf.write(tcl)
-            tcl_file = tf.name
-        try:
-            result = subprocess.run(
-                [xsct_path, tcl_file],
-                capture_output=True, text=True, timeout=timeout,
-            )
-            os.unlink(tcl_file)
-            out = (result.stdout + result.stderr).strip()
-            ok = result.returncode == 0 or any(
-                p in out for p in (
-                    "Download successful", "Done", "done",
-                    "Successfully downloaded",
-                )
-            )
-            return ok, out
-        except subprocess.TimeoutExpired:
-            os.unlink(tcl_file)
-            raise
-
-    # ── Attempt 1: no ps7_init (DAP alive from BIT step) ──
     if callback:
-        callback("progress", "Downloading ELF (DAP should be alive)...")
+        callback("progress", "Downloading FSBL to OCM...")
 
-    tcl1 = "\n".join([
+    tcl = "\n".join([
         "connect",
         "targets 2",
+        "dow \"%s\"" % fsbl_path,
+        "puts \"FSBL downloaded\"",
+        "con",
+        "after 3000",
+        "stop",
+        "puts \"FSBL done, CPU halted\"",
         "dow \"%s\"" % elf_path,
         "mwr 0xF800025C 0x00000001",
         "con",
+        "puts \"App running\"",
         "exit",
     ])
+
+    with tempfile.NamedTemporaryFile(
+        mode="w", suffix=".tcl", delete=False
+    ) as tf:
+        tf.write(tcl)
+        tcl_file = tf.name
+
     try:
-        ok1, out1 = _run(tcl1)
+        result = subprocess.run(
+            [xsct_path, tcl_file],
+            capture_output=True, text=True, timeout=timeout,
+        )
+        os.unlink(tcl_file)
+        output = (result.stdout + result.stderr).strip()
+
+        success = result.returncode == 0 or any(
+            p in output for p in (
+                "App running", "Download successful",
+                "Successfully downloaded",
+            )
+        )
+
+        if callback:
+            callback(
+                "complete" if success else "error",
+                "ELF " + ("executed via FSBL" if success else "failed"),
+            )
+
+        return BitstreamResult(
+            step="elf",
+            success=success,
+            message="ELF executed via FSBL" if success else "ELF execution failed",
+            output=output[:4000],
+        )
     except subprocess.TimeoutExpired:
+        os.unlink(tcl_file)
         return BitstreamResult(step="elf", success=False,
                                message="ELF download timed out")
 
-    if ok1:
-        if callback:
-            callback("complete", "ELF executed")
-        return BitstreamResult(step="elf", success=True,
-                               message="ELF executed successfully",
-                               output=out1[:4000])
 
-    # ── Attempt 2: re-init PS to recover DAP ──
-    if not has_ps7:
-        if callback:
-            callback("error", "ELF failed — no ps7_init.tcl for recovery")
-        return BitstreamResult(step="elf", success=False,
-                               message="ELF execution failed",
-                               output=out1[:4000])
+def _run_elf_direct(
+    xsct_path: str,
+    elf_path: Path,
+    callback: ProgressCallback = None,
+    timeout: int = 60,
+) -> BitstreamResult:
+    """Download and run ELF directly (no FSBL, assumes DAP alive).
+
+    Used when no FSBL ELF is configured.  Attempts direct dow
+    which only works if the CPU is in a clean state with MMU off.
+
+    Args:
+        xsct_path: Path to xsct executable.
+        elf_path: Path to .elf file.
+        callback: Optional progress callback.
+        timeout: Subprocess timeout in seconds.
+
+    Returns:
+        BitstreamResult with status.
+    """
+    import tempfile, os
 
     if callback:
-        callback(
-            "progress",
-            "DAP broken, re-initializing PS then retrying ELF...",
-        )
+        callback("progress", "Downloading ELF directly...")
 
-    tcl2 = "\n".join([
+    tcl = "\n".join([
         "connect",
-        "targets 1",
-        "source \"%s\"" % ps7_init_tcl,
-        "ps7_init",
         "targets 2",
         "dow \"%s\"" % elf_path,
         "mwr 0xF800025C 0x00000001",
         "con",
         "exit",
     ])
+
+    with tempfile.NamedTemporaryFile(
+        mode="w", suffix=".tcl", delete=False
+    ) as tf:
+        tf.write(tcl)
+        tcl_file = tf.name
+
     try:
-        ok2, out2 = _run(tcl2)
+        result = subprocess.run(
+            [xsct_path, tcl_file],
+            capture_output=True, text=True, timeout=timeout,
+        )
+        os.unlink(tcl_file)
+        output = (result.stdout + result.stderr).strip()
+
+        success = result.returncode == 0 or any(
+            p in output for p in (
+                "Download successful", "Successfully downloaded",
+            )
+        )
+
+        if callback:
+            callback("complete" if success else "error",
+                     "ELF " + ("executed" if success else "failed"))
+
+        return BitstreamResult(
+            step="elf",
+            success=success,
+            message="ELF executed" if success else "ELF execution failed",
+            output=output[:4000],
+        )
     except subprocess.TimeoutExpired:
+        os.unlink(tcl_file)
         return BitstreamResult(step="elf", success=False,
-                               message="ELF download (recovery) timed out")
-
-    if callback:
-        callback("complete" if ok2 else "error",
-                 "ELF " + ("executed (recovery)" if ok2 else "failed after recovery"))
-
-    return BitstreamResult(
-        step="elf",
-        success=ok2,
-        message=("ELF executed (DAP recovery)" if ok2 else "ELF execution failed"),
-        output=out2[:4000],
-    )
+                               message="ELF download timed out")
 
 
 # ── Full Workflow ───────────────────────────────────────────────────