Add wolfBoot FIT support for loading bitstream

Author: dgarske

arch.mk

@@ -74,6 +74,8 @@ ifeq ($(ARCH),AARCH64)
     # Support detection and skip of U-Boot legacy header */
     CFLAGS+=-DWOLFBOOT_UBOOT_LEGACY
     CFLAGS+=-DWOLFBOOT_DUALBOOT
+    # FPGA bitstream-from-FIT is supported via the PMU firmware
+    # (PM_FPGA_LOAD). Enable with FPGA_BITSTREAM=1 (opt-in).
 
     ifeq ($(HW_SHA3),1)
       # Use HAL for hash (see zynqmp.c)
@@ -90,6 +92,9 @@ ifeq ($(ARCH),AARCH64)
     CFLAGS+=-DWOLFBOOT_DUALBOOT
     # Support detection and skip of U-Boot legacy header
     CFLAGS+=-DWOLFBOOT_UBOOT_LEGACY
+    # NOTE: FPGA_BITSTREAM is stubbed on Versal (PL config is a PDI loaded
+    # by the PLM via XilLoader Load-PDI IPI - not yet implemented). Leave
+    # FPGA_BITSTREAM off until that path lands.
     # PLM owns RVBAR on Versal in JTAG boot; skip RVBAR writes
     CFLAGS+=-DSKIP_RVBAR=1
     # Disable SDMA for multi-block transfers - use PIO instead.
@@ -350,6 +355,8 @@ ifeq ($(ARCH),ARM)
      # positive probability from ~2^-32 to ~2^-64, matching what U-Boot's
      # own mkimage/bootm does.
      CFLAGS+=-DWOLFBOOT_UBOOT_LEGACY
+     # FPGA bitstream-from-FIT is supported via the DevC/PCAP DMA engine
+     # (full bitstream). Enable with FPGA_BITSTREAM=1 (opt-in).
   endif
 
   ifeq ($(TARGET),va416x0)

docs/Targets.md

@@ -4070,6 +4070,58 @@ FDT: Set chosen (...), linux,initrd-start=1073741824
 FDT: Set chosen (...), linux,initrd-end=...
 ```
 
+**FPGA bitstream from FIT**
+
+wolfBoot can program the PL (FPGA fabric) from an `fpga` sub-image carried in the same signed FIT, using the standard U-Boot convention: a sub-image with `type = "fpga"`, referenced from the configuration node via an optional `fpga = "<node>"` property, with a `compatible` string naming the load method. The PL is programmed before the kernel/DTB are loaded so PL-dependent clocks and peripherals come up first. The outer wolfBoot signature authenticates the whole FIT (bitstream included), so no per-image hashing is required.
+
+Enable with `FPGA_BITSTREAM=1` (off by default). A failed PL load is fatal (`wolfBoot_panic`) unless `FPGA_NONFATAL=1` is also set, which downgrades it to a logged warning that continues the boot.
+
+```sh
+cp config/examples/zynqmp.config .config
+make FPGA_BITSTREAM=1
+```
+
+Per-target support:
+- **ZynqMP** (`TARGET=zynq`): supported. The bitstream (a bootgen `.bin`, not `.bit`) is staged to its `load` address in DDR and handed to the PMU firmware via the `PM_FPGA_LOAD` EEMI call (xilfpga over the CSU DMA / PCAP).
+- **Zynq-7000** (`TARGET=zynq7000`): supported (full bitstream). Programmed directly through the DevC/PCAP DMA engine (UG585 ch.6). Partial reconfiguration is not yet implemented.
+- **Versal** (`TARGET=versal`): not yet implemented - Versal programs the PL with a PDI loaded by the PLM (XilLoader Load-PDI IPI), not a raw bitstream. `hal_fpga_load` is a stub; leave `FPGA_BITSTREAM` off on Versal until that path lands.
+
+The `compatible` string selects full vs partial: any value containing `partial` requests partial reconfiguration, otherwise a full bitstream is loaded. Typical full-bitstream values are `u-boot,zynqmp-fpga-ddrauth`, `u-boot,zynqmp-fpga-enc`, or `u-boot,fpga-legacy`.
+
+Example FIT layout:
+
+```dts
+images {
+    kernel-1 { ... };
+    fdt-1    { ... };
+    fpga-1 {
+        description = "FPGA bitstream";
+        data = /incbin/("system.bit.bin");  /* bootgen .bin */
+        type = "fpga";
+        arch = "arm64";                      /* "arm" for zynq7000 */
+        compression = "none";
+        load = <0x10000000>;                 /* DDR staging address */
+        compatible = "u-boot,zynqmp-fpga-ddrauth";
+        hash-1 { algo = "sha256"; };
+    };
+};
+configurations {
+    default = "conf-zcu102";
+    conf-zcu102 {
+        kernel = "kernel-1";
+        fdt    = "fdt-1";
+        fpga   = "fpga-1";
+    };
+};
+```
+
+Successful programming prints (ZynqMP):
+```
+FIT: programming FPGA 'fpga-1' (N bytes, full)
+FPGA status: 0x...
+FIT: FPGA programmed
+```
+
 
 ## Xilinx Zynq-7000 (ZC702)
 

hal/versal.c

@@ -1354,6 +1354,25 @@ int RAMFUNCTION hal_flash_erase(uintptr_t address, int len)
     return -1;
 }
 
+#ifdef WOLFBOOT_FPGA_BITSTREAM
+/* Versal programs the PL with a PDI (Programmable Device Image), not a raw
+ * bitstream. Runtime PL configuration is done by the PLM via the XilLoader
+ * "Load PDI" command (XLOADER_CMD_ID_LOAD_PDI), reached by building an IPI
+ * request to the PLM channel with the DDR PDI address/size and polling the
+ * PLM response. That path is not yet implemented; until it lands, leave
+ * FPGA_BITSTREAM disabled for Versal (the FIT loader treats a failed load
+ * as fatal unless FPGA_NONFATAL is set). */
+int hal_fpga_load(uint32_t flags, uintptr_t addr, size_t size)
+{
+    (void)flags;
+    (void)addr;
+    (void)size;
+    wolfBoot_printf("Versal FPGA/PDI load not implemented "
+        "(needs PLM XilLoader Load-PDI IPI)\n");
+    return -1;
+}
+#endif /* WOLFBOOT_FPGA_BITSTREAM */
+
 
 /* ============================================================================
  * External Flash Interface

hal/versal.its

@@ -29,13 +29,35 @@
                 algo = "sha256";
             };
         };
+        /* FPGA bitstream sub-image (requires FPGA_BITSTREAM=1). Add the
+         * "fpga = ..." reference to the configuration node below. The
+         * bitstream must be a bootgen .bin staged to its load address.
+         * NOTE: Versal PL config is a PDI loaded by the PLM (XilLoader
+         * Load-PDI IPI) - the wolfBoot Versal hal_fpga_load is currently
+         * a stub, so leave this disabled on Versal. On ZynqMP/Zynq-7000
+         * this is the working convention.
+         *
+         * fpga-1 {
+         *     description = "FPGA bitstream";
+         *     data = /incbin/("../system.bit.bin");
+         *     type = "fpga";
+         *     arch = "arm64";
+         *     compression = "none";
+         *     load = <0x10000000>;
+         *     compatible = "u-boot,zynqmp-fpga-ddrauth";
+         *     hash-1 {
+         *         algo = "sha256";
+         *     };
+         * };
+         */
     };
     configurations {
         default = "conf1";
         conf1 {
             description = "Linux kernel and FDT blob";
             kernel = "kernel-1";
             fdt = "fdt-1";
+            /* fpga = "fpga-1"; */
             hash-1 {
                 algo = "sha256";
             };

hal/zynq.c

@@ -197,6 +197,13 @@ static void smc_call(struct pt_regs *args)
 #define PM_MMIO_WRITE      0x13
 #define PM_MMIO_READ       0x14
 
+/* FPGA / PL programming (xilfpga via PMU firmware / TF-A) */
+#define PM_FPGA_LOAD       0x16 /* 22 */
+#define PM_FPGA_GET_STATUS 0x17 /* 23 */
+/* pm_fpga_load flags (bit 0 selects full vs partial bitstream) */
+#define XFPGA_FULLBIT_EN   0x0
+#define XFPGA_PARTIAL_EN   0x1
+
 /* AES */
 /* requires PMU built with -DENABLE_SECURE_VAL=1 */
 #define PM_SECURE_AES      0x2F
@@ -1713,6 +1720,43 @@ int RAMFUNCTION hal_flash_erase(uintptr_t address, int len)
     return 0;
 }
 
+#ifdef WOLFBOOT_FPGA_BITSTREAM
+/* Program the PL by handing the bitstream to the PMU firmware (xilfpga)
+ * via the PM_FPGA_LOAD EEMI call. The bitstream must be a bootgen .bin
+ * resident in DDR; it is flushed from the D-cache so the CSU DMA sees
+ * the committed bytes. */
+int hal_fpga_load(uint32_t flags, uintptr_t addr, size_t size)
+{
+    uint32_t ret_payload[PM_ARGS_CNT];
+    uint32_t pmflags;
+    uint32_t words;
+
+    /* PMU firmware expects the size as a count of 32-bit words. */
+    words = (uint32_t)((size + 3U) / 4U);
+    pmflags = (flags == HAL_FPGA_PARTIAL) ? XFPGA_PARTIAL_EN : XFPGA_FULLBIT_EN;
+
+    /* Ensure the bitstream is committed to DDR before the CSU DMA reads it. */
+    flush_dcache_range((unsigned long)addr, (unsigned long)(addr + size));
+
+    memset(ret_payload, 0, sizeof(ret_payload));
+    /* arg0=addr_low, arg1=addr_high, arg2=size(words), arg3=flags */
+    pmu_request(PM_FPGA_LOAD,
+        (uint32_t)(addr & 0xFFFFFFFF), (uint32_t)((uint64_t)addr >> 32),
+        words, pmflags, ret_payload);
+    if (ret_payload[0] != 0) {
+        wolfBoot_printf("PM_FPGA_LOAD failed: %u\n", ret_payload[0]);
+        return -1;
+    }
+
+    /* Confirm the PL reports configured (PCAP status). */
+    memset(ret_payload, 0, sizeof(ret_payload));
+    pmu_request(PM_FPGA_GET_STATUS, 0, 0, 0, 0, ret_payload);
+    wolfBoot_printf("FPGA status: 0x%x\n", ret_payload[1]);
+
+    return 0;
+}
+#endif /* WOLFBOOT_FPGA_BITSTREAM */
+
 /* Xilinx Write uses SPI mode and Page Program 0x02 */
 /* Issues using write with QSPI mode */
 int RAMFUNCTION ext_flash_write(uintptr_t address, const uint8_t *data, int len)

hal/zynq7000.c

@@ -871,6 +871,114 @@ uint64_t hal_get_timer_us(void)
     return (count * 1000000ULL) / (uint64_t)Z7_GTIMER_FREQ_HZ;
 }
 
+#ifdef WOLFBOOT_FPGA_BITSTREAM
+/* PL programming timeout (microseconds). */
+#ifndef Z7_FPGA_TIMEOUT_US
+#define Z7_FPGA_TIMEOUT_US 1000000ULL /* 1 second */
+#endif
+
+/* Clean the D-cache over [start, start+len) by MVA so the DevC DMA sees
+ * the committed bitstream bytes (DCCMVAC, L1 line = 32B). Mirrors the
+ * SDMA coherency path. */
+static void z7_dcache_clean_range(uintptr_t start, uint32_t len)
+{
+    uintptr_t addr;
+    uintptr_t end = (start + len + 31U) & ~31U;
+    start &= ~31U;
+    for (addr = start; addr < end; addr += 32U) {
+        __asm__ volatile("mcr p15, 0, %0, c7, c10, 1" : : "r"(addr) : "memory");
+    }
+    __asm__ volatile("dsb sy" : : : "memory");
+}
+
+/* Program the PL from a bootgen .bin bitstream resident in DDR using the
+ * DevC PCAP DMA engine (UG585 ch.6 / Xilinx XDcfg full-bitstream flow).
+ * Only the full-bitstream path is implemented; partial reconfiguration
+ * returns an error. */
+int hal_fpga_load(uint32_t flags, uintptr_t addr, size_t size)
+{
+    uint32_t sts;
+    uint32_t words = (uint32_t)((size + 3U) / 4U);
+    uint64_t t0;
+
+    if (flags == HAL_FPGA_PARTIAL) {
+        /* Partial reconfig leaves PROG_B asserted and routes via PCAP_PR;
+         * not implemented in this initial full-bitstream path. */
+        wolfBoot_printf("Z7 FPGA: partial reconfig not implemented\n");
+        return -1;
+    }
+    if (addr > 0xFFFFFFFFU || size == 0) {
+        return -1;
+    }
+
+    /* 1. Unlock DevC and select PCAP (not ICAP). */
+    Z7_DEVC_UNLOCK = Z7_DEVC_UNLOCK_KEY;
+    Z7_DEVC_CTRL |= (Z7_DEVC_CTRL_PCAP_MODE | Z7_DEVC_CTRL_PCAP_PR);
+    /* Disable internal PCAP loopback. */
+    Z7_DEVC_MCTRL &= ~Z7_DEVC_MCTRL_PCAP_LPBK;
+
+    /* 2. Clear sticky interrupts. */
+    Z7_DEVC_INT_STS = Z7_DEVC_INT_ALL;
+
+    /* 3. Pulse PROG_B low then high to clear the PL (full bitstream). */
+    Z7_DEVC_CTRL &= ~Z7_DEVC_CTRL_PCFG_PROG_B;
+    t0 = hal_get_timer_us();
+    while (Z7_DEVC_STATUS & Z7_DEVC_STATUS_PCFG_INIT) {
+        if (hal_get_timer_us() - t0 > Z7_FPGA_TIMEOUT_US) {
+            wolfBoot_printf("Z7 FPGA: timeout waiting PCFG_INIT clear\n");
+            return -1;
+        }
+    }
+    Z7_DEVC_CTRL |= Z7_DEVC_CTRL_PCFG_PROG_B;
+    t0 = hal_get_timer_us();
+    while (!(Z7_DEVC_STATUS & Z7_DEVC_STATUS_PCFG_INIT)) {
+        if (hal_get_timer_us() - t0 > Z7_FPGA_TIMEOUT_US) {
+            wolfBoot_printf("Z7 FPGA: timeout waiting PCFG_INIT set\n");
+            return -1;
+        }
+    }
+
+    /* 4. Clear interrupts and flush the config FIFOs. */
+    Z7_DEVC_INT_STS = Z7_DEVC_INT_ALL;
+    Z7_DEVC_MCTRL |= (Z7_DEVC_MCTRL_RFIFO_FLUSH | Z7_DEVC_MCTRL_WFIFO_FLUSH);
+
+    /* 5. Make the bitstream coherent in DDR for the DMA. */
+    z7_dcache_clean_range(addr, (uint32_t)size);
+
+    /* 6. Program the DMA: src = DDR bitstream (LSB=1 marks last descriptor),
+     *    dst = PCAP sentinel. Lengths are in 32-bit words. */
+    Z7_DEVC_DMA_SRC = ((uint32_t)addr) | Z7_DEVC_DMA_LAST;
+    Z7_DEVC_DMA_DST = Z7_DEVC_DMA_DEST_PCAP;
+    Z7_DEVC_DMA_SRC_LEN = words;
+    Z7_DEVC_DMA_DST_LEN = words;
+
+    /* 7. Wait for DMA done (and check error bits). */
+    t0 = hal_get_timer_us();
+    do {
+        sts = Z7_DEVC_INT_STS;
+        if (sts & Z7_DEVC_INT_ERR_MASK) {
+            wolfBoot_printf("Z7 FPGA: DMA error, INT_STS=0x%x\n", sts);
+            return -1;
+        }
+        if (hal_get_timer_us() - t0 > Z7_FPGA_TIMEOUT_US) {
+            wolfBoot_printf("Z7 FPGA: timeout waiting DMA done\n");
+            return -1;
+        }
+    } while (!(sts & Z7_DEVC_INT_DMA_DONE));
+
+    /* 8. Wait for the PL to report configuration complete. */
+    t0 = hal_get_timer_us();
+    while (!(Z7_DEVC_INT_STS & Z7_DEVC_INT_PCFG_DONE)) {
+        if (hal_get_timer_us() - t0 > Z7_FPGA_TIMEOUT_US) {
+            wolfBoot_printf("Z7 FPGA: timeout waiting PCFG_DONE\n");
+            return -1;
+        }
+    }
+
+    return 0;
+}
+#endif /* WOLFBOOT_FPGA_BITSTREAM */
+
 #if defined(DISK_SDCARD) || defined(DISK_EMMC)
 /* ============================================================================
  * SDHCI (SD Card / eMMC) Platform Support

hal/zynq7000.h

@@ -199,8 +199,51 @@
 #define Z7_GTIMER_FREQ_HZ   333333333UL
 #endif
 
-/* DevC (Device Configuration: AES + bitstream loader). UG585 ch.6. */
+/* DevC (Device Configuration: AES + bitstream loader). UG585 ch.6.
+ * Register offsets/bits mirror Xilinx xdevcfg_hw.h (XDCFG_*). */
 #define Z7_DEVC_BASE        0xF8007000UL
+#define Z7_DEVC_CTRL        (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x00)))
+#define Z7_DEVC_LOCK        (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x04)))
+#define Z7_DEVC_CFG         (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x08)))
+#define Z7_DEVC_INT_STS     (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x0C)))
+#define Z7_DEVC_INT_MASK    (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x10)))
+#define Z7_DEVC_STATUS      (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x14)))
+#define Z7_DEVC_DMA_SRC     (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x18)))
+#define Z7_DEVC_DMA_DST     (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x1C)))
+#define Z7_DEVC_DMA_SRC_LEN (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x20)))
+#define Z7_DEVC_DMA_DST_LEN (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x24)))
+#define Z7_DEVC_UNLOCK      (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x34)))
+#define Z7_DEVC_MCTRL       (*((volatile uint32_t*)(Z7_DEVC_BASE + 0x80)))
+
+/* CTRL bits */
+#define Z7_DEVC_CTRL_FORCE_RST     0x80000000U /* bit 31 */
+#define Z7_DEVC_CTRL_PCFG_PROG_B   0x40000000U /* bit 30 */
+#define Z7_DEVC_CTRL_PCAP_PR       0x08000000U /* bit 27: 1=PCAP, 0=ICAP */
+#define Z7_DEVC_CTRL_PCAP_MODE     0x04000000U /* bit 26: PCAP enable */
+#define Z7_DEVC_CTRL_QUARTER_RATE  0x02000000U /* bit 25 */
+
+/* INT_STS bits */
+#define Z7_DEVC_INT_DMA_DONE       0x00002000U /* bit 13 */
+#define Z7_DEVC_INT_D_P_DONE       0x00001000U /* bit 12: DMA+PCAP done */
+#define Z7_DEVC_INT_PCFG_DONE      0x00000004U /* bit 2 */
+/* DMA / AXI / config error aggregate (UG585 Table 6-x). */
+#define Z7_DEVC_INT_ERR_MASK       0x00F0C860U
+#define Z7_DEVC_INT_ALL            0xFFFFFFFFU
+
+/* STATUS bits */
+#define Z7_DEVC_STATUS_DMA_CMD_FULL 0x80000000U /* bit 31 */
+#define Z7_DEVC_STATUS_PCFG_INIT    0x00000010U /* bit 4 */
+
+/* MCTRL bits */
+#define Z7_DEVC_MCTRL_PCAP_LPBK    0x00000010U /* internal loopback */
+#define Z7_DEVC_MCTRL_RFIFO_FLUSH  0x00000400U /* bit 10 */
+#define Z7_DEVC_MCTRL_WFIFO_FLUSH  0x00000200U /* bit 9 */
+
+/* UNLOCK magic and PCAP DMA sentinel (UG585 ch.6 / XDcfg). The DMA
+ * address LSB set to 1 marks the last (only) descriptor in the chain. */
+#define Z7_DEVC_UNLOCK_KEY         0x757BDF0DU
+#define Z7_DEVC_DMA_DEST_PCAP      0xFFFFFFFFU
+#define Z7_DEVC_DMA_LAST           0x00000001U
 
 /* GIC (PL390 / GIC-400 v1) - per-CPU interface and distributor. */
 #define Z7_GIC_CPUIF_BASE   0xF8F00100UL