Philippe Mathieu-Daudé <phi...@linaro.org> 於 2025年5月8日 週四 下午2:19寫道：
>
> Hi Tim,
>
> On 8/5/25 04:15, Tim Lee wrote:
> > Fix flash device part number to `mx66l1g45g` according image-bmc run on
> > npcm8xx
> > evb board (SPIFlash...SF: Detected mx66l1g45g, total 128 MiB)
> >
> > And add auto zero flash image size to resolve error below after executing
> > `./qemu-system-aarch64 -machine npcm845-evb -drive file=image-bmc`
> >
> > Error message:
> > qemu-system-aarch64: mx66l1g45g device '/machine/unattached/device[73]'
> > requires 134217728 bytes, mtd0 block backend provides 67108864 bytes
> >
> > Tested:
> > Build passes and runs ./qemu-system-aarch64 -machine npcm845-evb normally
> >
> > Signed-off-by: Tim Lee <timlee660...@gmail.com>
> > ---
> > Changes since v1:
> > - Add a statement that checks whether the storage is writable
> >
> > hw/arm/npcm8xx_boards.c | 20 +++++++++++++++++++-
> > 1 file changed, 19 insertions(+), 1 deletion(-)
> >
> > diff --git a/hw/arm/npcm8xx_boards.c b/hw/arm/npcm8xx_boards.c
> > index 3fb8478e72..79295a586c 100644
> > --- a/hw/arm/npcm8xx_boards.c
> > +++ b/hw/arm/npcm8xx_boards.c
> > @@ -27,6 +27,7 @@
> > #include "qemu/error-report.h"
> > #include "qemu/datadir.h"
> > #include "qemu/units.h"
> > +#include "system/block-backend.h"
> >
> > #define NPCM845_EVB_POWER_ON_STRAPS 0x000017ff
> >
> > @@ -59,10 +60,26 @@ static void npcm8xx_connect_flash(NPCM7xxFIUState *fiu,
> > int cs_no,
> > {
> > DeviceState *flash;
> > qemu_irq flash_cs;
> > + BlockBackend *blk;
> > + BlockDriverState *bs;
> > + uint64_t blk_size, perm, shared_perm;
> >
> > flash = qdev_new(flash_type);
> > if (dinfo) {
> > qdev_prop_set_drive(flash, "drive", blk_by_legacy_dinfo(dinfo));
> > + blk = blk_by_legacy_dinfo(dinfo);
> > + bs = blk_bs(blk);
> > + blk_size = blk_getlength(blk);
> > +
> > + if (!bdrv_is_read_only(bs)) {
>
> This isn't what I meant, we'll get the same issue with read-only storage.
>
> See:
> https://lore.kernel.org/qemu-devel/cafeaca9itedtrznx1krveza__dch95abppzbdtj0-tuxwih...@mail.gmail.com/
>
> > + if (blk_size < fiu->flash_size) {
> > + blk_get_perm(blk, &perm, &shared_perm);
> > + blk_set_perm(blk, BLK_PERM_ALL, BLK_PERM_ALL,
> > &error_abort);
> > + blk_truncate(blk, fiu->flash_size, true, PREALLOC_MODE_OFF,
> > + BDRV_REQ_ZERO_WRITE, &error_abort);
> > + blk_set_perm(blk, perm, shared_perm, &error_abort);
> > + }
> > + }
> > }
> > qdev_realize_and_unref(flash, BUS(fiu->spi), &error_fatal);
> >
> > @@ -194,7 +211,8 @@ static void npcm845_evb_init(MachineState *machine)
> > qdev_realize(DEVICE(soc), NULL, &error_fatal);
> >
> > npcm8xx_load_bootrom(machine, soc);
> > - npcm8xx_connect_flash(&soc->fiu[0], 0, "w25q256", drive_get(IF_MTD, 0,
> > 0));
> > + npcm8xx_connect_flash(&soc->fiu[0], 0, "mx66l1g45g",
> > + drive_get(IF_MTD, 0, 0));
> > npcm845_evb_i2c_init(soc);
> > npcm845_evb_fan_init(NPCM8XX_MACHINE(machine), soc);
> > npcm8xx_load_kernel(machine, soc);
>
Hi Philippe,
Thanks for your sharing. Now I understand why you say it won't work on
read-only storage.
Currently, without this change we cannot run the npcm845-evb of QEMU
with a large flash size (128 MB), including the latest OpenBMC build
image.
That's why we sync the method into npcm8xx_board for our npcm845 evb
board. It's seems not suitable for upstream, right? Thanks.
--
Best regards,
Tim Lee
