[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [PATCH 4/5] hw/arm/aspeed: add a 'execute-in-place' property to boot
From: |
Cédric Le Goater |
Subject: |
Re: [PATCH 4/5] hw/arm/aspeed: add a 'execute-in-place' property to boot directly from CE0 |
Date: |
Fri, 10 Jan 2020 11:14:43 +0100 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Thunderbird/68.2.2 |
On 1/7/20 9:34 AM, Philippe Mathieu-Daudé wrote:
> On 1/7/20 8:34 AM, Cédric Le Goater wrote:
>> The overhead for the OpenBMC firmware images using the a custom U-Boot
>> is around 2 seconds, which is fine, but with a U-Boot from mainline,
>> it takes an extra 50 seconds or so to reach Linux. A quick survey on
>> the number of reads performed on the flash memory region gives the
>> following figures :
>>
>> OpenBMC U-Boot 922478 (~ 3.5 MBytes)
>> Mainline U-Boot 20569977 (~ 80 MBytes)
>>
>> QEMU must be trashing the TCG TBs and reloading text very often. Some
>> addresses are read more than 250.000 times. Until we find a solution
>> to improve boot time, execution from MMIO is not activated by default.
>>
>> Setting this option also breaks migration compatibility.
>>
>> Signed-off-by: Cédric Le Goater <address@hidden>
>> Reviewed-by: Peter Maydell <address@hidden>
>> ---
>> include/hw/arm/aspeed.h | 2 ++
>> hw/arm/aspeed.c | 44 ++++++++++++++++++++++++++++++++++++-----
>> 2 files changed, 41 insertions(+), 5 deletions(-)
>>
>> diff --git a/include/hw/arm/aspeed.h b/include/hw/arm/aspeed.h
>> index 4423cd0cda71..18521484b90e 100644
>> --- a/include/hw/arm/aspeed.h
>> +++ b/include/hw/arm/aspeed.h
>> @@ -19,6 +19,8 @@ typedef struct AspeedBoardState AspeedBoardState;
>> typedef struct AspeedMachine {
>> MachineState parent_obj;
>> +
>> + bool mmio_exec;
>> } AspeedMachine;
>> #define ASPEED_MACHINE_CLASS(klass) \
>> diff --git a/hw/arm/aspeed.c b/hw/arm/aspeed.c
>> index 0a7dfd29868b..bf23579fa53d 100644
>> --- a/hw/arm/aspeed.c
>> +++ b/hw/arm/aspeed.c
>> @@ -260,11 +260,18 @@ static void aspeed_machine_init(MachineState *machine)
>> * SoC and 128MB for the AST2500 SoC, which is twice as big as
>> * needed by the flash modules of the Aspeed machines.
>> */
>> - memory_region_init_rom(boot_rom, OBJECT(bmc), "aspeed.boot_rom",
>> - fl->size, &error_abort);
>> - memory_region_add_subregion(get_system_memory(), FIRMWARE_ADDR,
>> - boot_rom);
>> - write_boot_rom(drive0, FIRMWARE_ADDR, fl->size, &error_abort);
>> + if (ASPEED_MACHINE(machine)->mmio_exec) {
>> + memory_region_init_alias(boot_rom, OBJECT(bmc),
>> "aspeed.boot_rom",
>> + &fl->mmio, 0, fl->size);
>> + memory_region_add_subregion(get_system_memory(), FIRMWARE_ADDR,
>> + boot_rom);
>> + } else {
>> + memory_region_init_rom(boot_rom, OBJECT(bmc), "aspeed.boot_rom",
>> + fl->size, &error_abort);
>> + memory_region_add_subregion(get_system_memory(), FIRMWARE_ADDR,
>> + boot_rom);
>> + write_boot_rom(drive0, FIRMWARE_ADDR, fl->size, &error_abort);
>> + }
>
> Nitpick: I'd use rom_add_file_mr() in write_boot_rom(), and keep the
> memory_region_add_subregion() call after the if/else statement.
>
> Reviewed-by: Philippe Mathieu-Daudé <address@hidden>
I would rather address changes in follow-up patches.
Thanks for the hint.
C.
>
>> }
>> aspeed_board_binfo.ram_size = ram_size;
>> @@ -398,6 +405,30 @@ static void witherspoon_bmc_i2c_init(AspeedBoardState
>> *bmc)
>> /* Bus 11: TODO ucd90160@64 */
>> }
>> +static bool aspeed_get_mmio_exec(Object *obj, Error **errp)
>> +{
>> + return ASPEED_MACHINE(obj)->mmio_exec;
>> +}
>> +
>> +static void aspeed_set_mmio_exec(Object *obj, bool value, Error **errp)
>> +{
>> + ASPEED_MACHINE(obj)->mmio_exec = value;
>> +}
>> +
>> +static void aspeed_machine_instance_init(Object *obj)
>> +{
>> + ASPEED_MACHINE(obj)->mmio_exec = false;
>> +}
>> +
>> +static void aspeed_machine_class_props_init(ObjectClass *oc)
>> +{
>> + object_class_property_add_bool(oc, "execute-in-place",
>> + aspeed_get_mmio_exec,
>> + aspeed_set_mmio_exec, &error_abort);
>> + object_class_property_set_description(oc, "execute-in-place",
>> + "boot directly from CE0 flash device",
>> &error_abort);
>> +}
>> +
>> static void aspeed_machine_class_init(ObjectClass *oc, void *data)
>> {
>> MachineClass *mc = MACHINE_CLASS(oc);
>> @@ -407,6 +438,8 @@ static void aspeed_machine_class_init(ObjectClass *oc,
>> void *data)
>> mc->no_floppy = 1;
>> mc->no_cdrom = 1;
>> mc->no_parallel = 1;
>> +
>> + aspeed_machine_class_props_init(oc);
>> }
>> static void aspeed_machine_palmetto_class_init(ObjectClass *oc, void
>> *data)
>> @@ -549,6 +582,7 @@ static const TypeInfo aspeed_machine_types[] = {
>> .name = TYPE_ASPEED_MACHINE,
>> .parent = TYPE_MACHINE,
>> .instance_size = sizeof(AspeedMachine),
>> + .instance_init = aspeed_machine_instance_init,
>> .class_size = sizeof(AspeedMachineClass),
>> .class_init = aspeed_machine_class_init,
>> .abstract = true,
>>
>
- Re: [PATCH 2/5] hw/arm: ast2600: Wire up the eMMC controller, (continued)
[PATCH 3/5] ftgmac100: check RX and TX buffer alignment, Cédric Le Goater, 2020/01/07
[PATCH 4/5] hw/arm/aspeed: add a 'execute-in-place' property to boot directly from CE0, Cédric Le Goater, 2020/01/07
[PATCH 5/5] misc/pca9552: Add qom set and get, Cédric Le Goater, 2020/01/07
Re: [PATCH 0/5] aspeed: extensions and fixes, Cédric Le Goater, 2020/01/13