Hi Jassi
On Mon, 18 Jul 2022 at 18:34, Jassi Brar jaswinder.singh@linaro.org wrote:
On Mon, 18 Jul 2022 at 10:31, Jassi Brar jaswinder.singh@linaro.org wrote:
Hi Ilias,
On Mon, 18 Jul 2022 at 10:16, Ilias Apalodimas ilias.apalodimas@linaro.org wrote:
Hi Jassi
On Mon, 18 Jul 2022 at 18:08, Jassi Brar jaswinder.singh@linaro.org wrote:
On Mon, 18 Jul 2022 at 09:47, Ilias Apalodimas ilias.apalodimas@linaro.org wrote:
Hi all,
On Mon, 18 Jul 2022 at 17:43, Jassi Brar jaswinder.singh@linaro.org wrote:
On Mon, 20 Jun 2022 at 03:23, Michal Simek monstr@monstr.eu wrote: > On 6/9/22 14:30, Sughosh Ganu wrote: > > From: Masami Hiramatsu masami.hiramatsu@linaro.org > > ....
> > +} > > + > > +static int plat_sf_get_flash(struct spi_flash **flash) > > +{ > > + int ret = 0; > > + > > + if (!plat_spi_flash) > > + ret = __plat_sf_get_flash(); > > + > > + *flash = plat_spi_flash; > > + > > + return ret; > > +} > > + > > +static int sf_load_data(u32 offs, u32 size, void **data) > > +{ > > + struct spi_flash *flash; > > + int ret; > > + > > + ret = plat_sf_get_flash(&flash); > > + if (ret < 0) > > + return ret; > > + > > + *data = memalign(ARCH_DMA_MINALIGN, size); > > + if (!*data) > > + return -ENOMEM; > > + > > + ret = spi_flash_read(flash, offs, size, *data); > > + if (ret < 0) { > > + free(*data); > > + *data = NULL; > > + } > > + > > + return ret; > > +} > > + > > +static int sf_save_data(u32 offs, u32 size, void *data) > > +{ > > + struct spi_flash *flash; > > + u32 sect_size, nsect; > > + void *buf; > > + int ret; > > + > > + ret = plat_sf_get_flash(&flash); > > + if (ret < 0) > > + return ret; > > + > > + sect_size = flash->mtd.erasesize; > > + nsect = DIV_ROUND_UP(size, sect_size); > > + ret = spi_flash_erase(flash, offs, nsect * sect_size); > > What it is interesting here that framework itself is using mtd infrastructure > but this platform driver is calling spi functions directly. > It looks a little bit nonstandard way. What's the reason for it? > Yup, this whole sf shebang is unnecessary, and removed for next revision.
> > + > > +#define PLAT_METADATA_OFFSET 0x510000 > > +#define PLAT_METADATA_SIZE (sizeof(struct devbox_metadata)) > > + > > +struct __packed devbox_metadata { > > + u32 boot_index; > > + u32 boot_count; > > There is the whole bootcount infrastructure for this. I think it would be much > better to use that framework instead of creating parallel one. > Yes, this goes too.
Is bootcount really suited for this case? AFAIK bootcount either requires device specific registers (which won't reset on reboots), or an environment you can write data to. But what if a user wants to disable writing the env variables and the device doesn't have a set of registers we can use?
Maybe it should be moved in 'struct fwu_mdata' ?
I was mostly thinking on moving this count as another 'bootcount' method. So in case the user has disabled writing evn variables but he is booting with EFI he can use that.
Sorry, not sure I understand.... IIUIC there has to be some persistent storage.
Of the three options - registers, efi-env and mdata, I think the last one is more robust. For ex, if BL33 isn't reached after an update. We want BL2 (which may not have access to efi variables) to be able to revert the active index.
and which requires a bootcount for each stage. hmm... probably I am overlooking something.
Well it's indeed more complicated, but the reasoning was something along the lines of - What if BL2 crashes really early, before it can access storage? - BL2 doesn't have code to write that data only read it (in some cases, depends on how the data is stored)
So the solution was to have individual counters
Cheers /Ilias