On a real device, the boot ROM contains the very first instructions the CPU
executes. Also, U-Boot calls into the ROM to determine the boot device. While
we're not actually implementing this here, let's create the infrastructure and
add a dummy ROM with all zeros. This allows for implementing a ROM later without
touching the source code and even allows for users to provide their own ROMs.

The imx8mp-boot.rom was created with
`dd if=/dev/zero of=imx8mp-boot.rom bs=1 count=258048`.

Signed-off-by: Bernhard Beschow <shentey@gmail.com>
---
 MAINTAINERS                 |   1 +
 include/hw/arm/fsl-imx8mp.h |   1 +
 hw/arm/fsl-imx8mp.c         |  18 ++++++++++++++++++
 pc-bios/imx8mp-boot.rom     | Bin 0 -> 258048 bytes
 pc-bios/meson.build         |   1 +
 5 files changed, 21 insertions(+)
 create mode 100644 pc-bios/imx8mp-boot.rom

diff --git a/MAINTAINERS b/MAINTAINERS
index 2e7fc6fa91..489e426d85 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -831,6 +831,7 @@ F: hw/pci-host/fsl_imx8m_phy.c
 F: include/hw/arm/fsl-imx8mp.h
 F: include/hw/misc/imx8mp_*.h
 F: include/hw/pci-host/fsl_imx8m_phy.h
+F: pc-bios/imx8mp*
 F: docs/system/arm/imx8mp-evk.rst
 
 MPS2 / MPS3
diff --git a/include/hw/arm/fsl-imx8mp.h b/include/hw/arm/fsl-imx8mp.h
index 5247e972b8..4dbe30f524 100644
--- a/include/hw/arm/fsl-imx8mp.h
+++ b/include/hw/arm/fsl-imx8mp.h
@@ -68,6 +68,7 @@ struct FslImx8mpState {
     DesignwarePCIEHost pcie;
     FslImx8mPciePhyState   pcie_phy;
     OrIRQState         gpt5_gpt6_irq;
+    MemoryRegion       boot_rom;
 
     uint32_t           phy_num;
     bool               phy_connected;
diff --git a/hw/arm/fsl-imx8mp.c b/hw/arm/fsl-imx8mp.c
index f0d5980300..f26cf5984e 100644
--- a/hw/arm/fsl-imx8mp.c
+++ b/hw/arm/fsl-imx8mp.c
@@ -9,12 +9,14 @@
  */
 
 #include "qemu/osdep.h"
+#include "qemu/datadir.h"
 #include "exec/address-spaces.h"
 #include "hw/arm/bsa.h"
 #include "hw/arm/fsl-imx8mp.h"
 #include "hw/intc/arm_gicv3.h"
 #include "hw/misc/unimp.h"
 #include "hw/boards.h"
+#include "hw/loader.h"
 #include "system/system.h"
 #include "target/arm/cpu-qom.h"
 #include "qapi/error.h"
@@ -263,6 +265,7 @@ static void fsl_imx8mp_realize(DeviceState *dev, Error **errp)
     MachineState *ms = MACHINE(qdev_get_machine());
     FslImx8mpState *s = FSL_IMX8MP(dev);
     DeviceState *gicdev = DEVICE(&s->gic);
+    g_autofree char *filename = NULL;
     int i;
 
     if (ms->smp.cpus > FSL_IMX8MP_NUM_CPUS) {
@@ -644,10 +647,25 @@ static void fsl_imx8mp_realize(DeviceState *dev, Error **errp)
     sysbus_mmio_map(SYS_BUS_DEVICE(&s->pcie_phy), 0,
                     fsl_imx8mp_memmap[FSL_IMX8MP_PCIE_PHY1].addr);
 
+    /* ROM memory */
+    if (!memory_region_init_rom(&s->boot_rom, OBJECT(dev),
+                                fsl_imx8mp_memmap[FSL_IMX8MP_BOOT_ROM].name,
+                                fsl_imx8mp_memmap[FSL_IMX8MP_BOOT_ROM].size,
+                                errp)) {
+        return;
+    }
+    filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "imx8mp-boot.rom");
+    load_image_size(filename, memory_region_get_ram_ptr(&s->boot_rom),
+                    memory_region_size(&s->boot_rom));
+    memory_region_add_subregion(get_system_memory(),
+                                fsl_imx8mp_memmap[FSL_IMX8MP_BOOT_ROM].addr,
+                                &s->boot_rom);
+
     /* Unimplemented devices */
     for (i = 0; i < ARRAY_SIZE(fsl_imx8mp_memmap); i++) {
         switch (i) {
         case FSL_IMX8MP_ANA_PLL:
+        case FSL_IMX8MP_BOOT_ROM:
         case FSL_IMX8MP_CCM:
         case FSL_IMX8MP_GIC_DIST:
         case FSL_IMX8MP_GIC_REDIST:
diff --git a/pc-bios/imx8mp-boot.rom b/pc-bios/imx8mp-boot.rom
new file mode 100644
index 0000000000000000000000000000000000000000..5324b5eed200e723d048f8476e4d96d45622fd4d
GIT binary patch
literal 258048
zcmeIuF#!Mo0K%a4Pi+Q&h(KY$fB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM
z7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*
z1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd
z0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwA
zz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEj
zFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r
z3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@
z0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VK
zfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5
zV8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM
z7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*
z1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd
z0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwA
zz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEj
zFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r
z3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@
z0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VK
zfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5
zV8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM
z7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*
z1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd
z0RsjM7%*VKfB^#r3>YwAz<>b*1`HT5V8DO@0|pEjFkrxd0RsjM7%*VKfB^#r3>f$Z
E2JkHa0RR91

literal 0
HcmV?d00001

diff --git a/pc-bios/meson.build b/pc-bios/meson.build
index 51e95cc903..50506a7895 100644
--- a/pc-bios/meson.build
+++ b/pc-bios/meson.build
@@ -60,6 +60,7 @@ blobs = [
   'efi-virtio.rom',
   'efi-e1000e.rom',
   'efi-vmxnet3.rom',
+  'imx8mp-boot.rom',
   'qemu-nsis.bmp',
   'multiboot.bin',
   'multiboot_dma.bin',
-- 
2.48.1

On Sun, 23 Feb 2025 at 11:47, Bernhard Beschow <shentey@gmail.com> wrote:
>
> On a real device, the boot ROM contains the very first instructions the CPU
> executes. Also, U-Boot calls into the ROM to determine the boot device. While
> we're not actually implementing this here, let's create the infrastructure and
> add a dummy ROM with all zeros. This allows for implementing a ROM later without
> touching the source code and even allows for users to provide their own ROMs.
>
> The imx8mp-boot.rom was created with
> `dd if=/dev/zero of=imx8mp-boot.rom bs=1 count=258048`.
>
> Signed-off-by: Bernhard Beschow <shentey@gmail.com>

Hi; I've thought about this a bit, and I think we should just
postpone having code to load a rom blob until we have an
actual rom blob that's not all zeroes. This makes this
board the same as e.g. imx7 which also doesn't try to load
a ROM blob. You can create a memory region for the ROM,
like imx7; just don't load anything into it, and that's
exactly the same as if there was an all-zeroes file that
got loaded.

This doesn't prevent users from experimenting with providing
their own ROM code, because they can do that with the
generic loader (passing it either an ELF file set up to
load to the ROM area or else a binary blob plus the
start address of 0).

That way we aren't carrying code we aren't using and an
awkward binary blob in git; we can add those when we need
them.

thanks
-- PMM

Am 25. Februar 2025 14:29:15 UTC schrieb Peter Maydell <peter.maydell@linaro.org>:
>On Sun, 23 Feb 2025 at 11:47, Bernhard Beschow <shentey@gmail.com> wrote:
>>
>> On a real device, the boot ROM contains the very first instructions the CPU
>> executes. Also, U-Boot calls into the ROM to determine the boot device. While
>> we're not actually implementing this here, let's create the infrastructure and
>> add a dummy ROM with all zeros. This allows for implementing a ROM later without
>> touching the source code and even allows for users to provide their own ROMs.
>>
>> The imx8mp-boot.rom was created with
>> `dd if=/dev/zero of=imx8mp-boot.rom bs=1 count=258048`.
>>
>> Signed-off-by: Bernhard Beschow <shentey@gmail.com>
>
>Hi; I've thought about this a bit, and I think we should just
>postpone having code to load a rom blob until we have an
>actual rom blob that's not all zeroes. This makes this
>board the same as e.g. imx7 which also doesn't try to load
>a ROM blob. You can create a memory region for the ROM,
>like imx7; just don't load anything into it, and that's
>exactly the same as if there was an all-zeroes file that
>got loaded.
>
>This doesn't prevent users from experimenting with providing
>their own ROM code, because they can do that with the
>generic loader (passing it either an ELF file set up to
>load to the ROM area or else a binary blob plus the
>start address of 0).
>
>That way we aren't carrying code we aren't using and an
>awkward binary blob in git; we can add those when we need
>them.

Yes, makes sense.

Best regards,
Bernhard

>
>thanks
>-- PMM