Hi Ulf,
> -----Original Message-----
> From: Ulf Hansson [mailto:ulf.hans...@linaro.org]
> Sent: Wednesday, August 13, 2014 3:47 PM
> To: Avi Shchislowski; Grant Grundler; Alex Lemberg
> Cc: linux-mmc@vger.kernel.org; c...@laptop.org
> Subject: Re: [RFC PATCH 1/1 v8]mmc: Support-FFU-for-eMMC-v5.0
>
> On 16 July 2014 17:47, Avi Shchislowski <avi.shchislow...@sandisk.com>
> wrote:
> > The Field Firmware Update (FFU) feature is new for eMMC 5.0 spec (Jedec:
> JESD84-B50.pdf)
> >
> > http://www.jedec.org/standards-documents/technology-focus-
> areas/flash-
> > memory-ssds-ufs-emmc/e-mmc
> >
> > An ioctl has been added to provide the new firmware image's file name to
> the mmc driver and udev is then used to retrieve its data.
> >
> > Two new ioctls have been added:
> > 1. FFU download firmware - transfer the new firmware data from user
> > space to the eMMC device 2. FFU install - initializes the new firmware
> > update
> >
> >
> > Signed-off-by: Avi Shchislowski <avi.shchislow...@sandisk.com>
> > Signed-off-by: Alex Lemberg <alex.lemb...@sandisk.com>
>
> Hi guys,
>
> Sorry for the delay in providing feeback, but hey better late than never. :-)
Thanks, we appreciate your comments!
>
> >
> > ---
> > V8:
> > - Modified according to Gwendal Grignou comments and patch:
> > [PATCH] Fix on top of sandisk patches
> >
> > V7:
> > - fixed mangled white space
> >
> > V5:
> > - provides udev (request_firmware) implementation as advised in patch
> > v2 comments.
> >
> > diff --git a/drivers/mmc/card/Kconfig b/drivers/mmc/card/Kconfig index
> > 5562308..19ba729 100644
> > --- a/drivers/mmc/card/Kconfig
> > +++ b/drivers/mmc/card/Kconfig
> > @@ -68,3 +68,11 @@ config MMC_TEST
> >
> > This driver is only of interest to those developing or
> > testing a host driver. Most people should say N here.
> > +
> > +config MMC_FFU
> > + bool "FFU SUPPORT"
> > + depends on MMC != n
> > + help
> > + This is an option to run firmware update on eMMC 5.0.
> > + Field firmware updates (FFU) enables features enhancment
> > + in the field.
>
> Why do we need a new Kconfig for this? I think we can drop it.
The motivation was to let vendor decide if FFU code should be included in his
release or not.
Some vendors prefer using this FFU functionality during
production/testing/development only.
>
> > diff --git a/drivers/mmc/card/Makefile b/drivers/mmc/card/Makefile
> > index c73b406..1e9223b 100644
> > --- a/drivers/mmc/card/Makefile
> > +++ b/drivers/mmc/card/Makefile
> > @@ -8,3 +8,4 @@ obj-$(CONFIG_MMC_TEST) += mmc_test.o
> >
> > obj-$(CONFIG_SDIO_UART) += sdio_uart.o
> >
> > +obj-$(CONFIG_MMC_FFU) += ffu.o
> > diff --git a/drivers/mmc/card/block.c b/drivers/mmc/card/block.c index
> > 7b5424f..3ed900b 100644
> > --- a/drivers/mmc/card/block.c
> > +++ b/drivers/mmc/card/block.c
> > @@ -41,6 +41,7 @@
> > #include <linux/mmc/host.h>
> > #include <linux/mmc/mmc.h>
> > #include <linux/mmc/sd.h>
> > +#include <linux/mmc/ffu.h>
> >
> > #include <asm/uaccess.h>
> >
> > @@ -523,8 +524,18 @@ static int mmc_blk_ioctl_cmd(struct block_device
> > *bdev,
> >
> > mrq.cmd = &cmd;
> >
> > + if (cmd.opcode == MMC_FFU_DOWNLOAD_OP) {
> > + err = mmc_ffu_download(card, idata->buf);
> > + goto cmd_done;
> > + }
> > +
> > mmc_get_card(card);
> >
> > + if (cmd.opcode == MMC_FFU_INSTALL_OP) {
> > + err = mmc_ffu_install(card);
> > + goto cmd_rel_host;
> > + }
> > +
>
> I think one FFU operation code should be enough. Overall that would also
> simplify management and thus code.
The reason for splitting it to two operation codes is providing flexibility to
hosts
due to different implementation of power_off routine.
Some mmc hosts could not support suspend/resume, so this split will allow to
make
"manual" power reset instead of running MMC_FFU_INSTALL_OP. (JEDEC spec allows
doing power cycle).
>
> Additionally I think it's better to move these "if statements" prior a
> potential
> data transfer is set up. You don't use the mrq struct anyway.
Agree
>
> > err = mmc_blk_part_switch(card, md);
> > if (err)
> > goto cmd_rel_host;
> > diff --git a/drivers/mmc/card/ffu.c b/drivers/mmc/card/ffu.c new file
> > mode 100644 index 0000000..783673e
> > --- /dev/null
> > +++ b/drivers/mmc/card/ffu.c
>
> I am wondering whether this code should belong to
> drivers/mmc/core/mmc.c instead. Or unless it becomes too big, we could
> have a separate file called mmc_ffu.[c|h], but in the core directory.
For easiest maintenance we prefer using separate file.
mmc_ffu.[c|h] in core directory - will do.
>
> > @@ -0,0 +1,607 @@
> > +/*
> > + * * ffu.c
> > + *
> > + * Copyright 2007-2008 Pierre Ossman
> > + *
> > + * Modified by SanDisk Corp., Copyright (c) 2013 SanDisk Corp.
> > + *
> > + * This program is free software; you can redistribute it and/or
> > +modify
> > + * it under the terms of the GNU General Public License as published
> > +by
> > + * the Free Software Foundation; either version 2 of the License, or
> > +(at
> > + * your option) any later version.
> > + *
> > + * This program includes bug.h, card.h, host.h, mmc.h, scatterlist.h,
> > + * slab.h, ffu.h & swap.h header files
> > + * The original, unmodified version of this program - the mmc_test.c
> > + * file - is obtained under the GPL v2.0 license that is available
> > +via
> > + * http://www.gnu.org/licenses/,
> > + * or http://www.opensource.org/licenses/gpl-2.0.php
> > +*/
> > +
> > +#include <linux/bug.h>
> > +#include <linux/errno.h>
> > +#include <linux/mmc/card.h>
> > +#include <linux/mmc/host.h>
> > +#include <linux/mmc/mmc.h>
> > +#include <linux/scatterlist.h>
> > +#include <linux/slab.h>
> > +#include <linux/swap.h>
> > +#include <linux/mmc/ffu.h>
> > +#include <linux/firmware.h>
> > +
> > +/**
> > + * struct mmc_ffu_pages - pages allocated by 'alloc_pages()'.
> > + * @page: first page in the allocation
> > + * @order: order of the number of pages allocated */ struct
> > +mmc_ffu_pages {
> > + struct page *page;
> > + unsigned int order;
> > +};
> > +
> > +/**
> > + * struct mmc_ffu_mem - allocated memory.
> > + * @arr: array of allocations
> > + * @cnt: number of allocations
> > + */
> > +struct mmc_ffu_mem {
> > + struct mmc_ffu_pages *arr;
> > + unsigned int cnt;
> > +};
> > +
> > +struct mmc_ffu_area {
> > + unsigned long max_sz;
> > + unsigned int max_tfr;
> > + unsigned int max_segs;
> > + unsigned int max_seg_sz;
> > + unsigned int blocks;
> > + unsigned int sg_len;
> > + struct mmc_ffu_mem *mem;
> > + struct scatterlist *sg;
> > +};
> > +
> > +static void mmc_ffu_prepare_mrq(struct mmc_card *card,
> > + struct mmc_request *mrq, struct scatterlist *sg, unsigned int
> > sg_len,
> > + u32 arg, unsigned int blocks, unsigned int blksz) {
> > + BUG_ON(!mrq || !mrq->cmd || !mrq->data || !mrq->stop);
> > +
> > + if (blocks > 1)
> > + mrq->cmd->opcode = MMC_WRITE_MULTIPLE_BLOCK;
> > + else
> > + mrq->cmd->opcode = MMC_WRITE_BLOCK;
> > +
> > + mrq->cmd->arg = arg;
> > + if (!mmc_card_blockaddr(card))
> > + mrq->cmd->arg <<= 9;
> > +
> > + mrq->cmd->flags = MMC_RSP_R1 | MMC_CMD_ADTC;
> > + if (blocks == 1) {
> > + mrq->stop = NULL;
> > + } else {
> > + mrq->stop->opcode = MMC_STOP_TRANSMISSION;
> > + mrq->stop->arg = 0;
> > + mrq->stop->flags = MMC_RSP_R1B | MMC_CMD_AC;
> > + }
> > +
> > + mrq->data->blksz = blksz;
> > + mrq->data->blocks = blocks;
> > + mrq->data->flags = MMC_DATA_WRITE;
> > + mrq->data->sg = sg;
> > + mrq->data->sg_len = sg_len;
> > +
> > + mmc_set_data_timeout(mrq->data, card); }
> > +
> > +/*
> > + * Checks that a normal transfer didn't have any errors */ static
> > +int mmc_ffu_check_result(struct mmc_request *mrq) {
> > + BUG_ON(!mrq || !mrq->cmd || !mrq->data);
> > +
> > + if (mrq->cmd->error != 0)
> > + return -EINVAL;
> > +
> > + if (mrq->data->error != 0)
> > + return -EINVAL;
> > +
> > + if (mrq->stop != NULL && mrq->stop->error != 0)
> > + return -1;
> > +
> > + if (mrq->data->bytes_xfered != (mrq->data->blocks * mrq->data-
> >blksz))
> > + return -EINVAL;
> > +
> > + return 0;
> > +}
> > +
> > +static int mmc_ffu_busy(struct mmc_command *cmd) {
> > + return !(cmd->resp[0] & R1_READY_FOR_DATA) ||
> > + (R1_CURRENT_STATE(cmd->resp[0]) == R1_STATE_PRG); }
> > +
> > +static int mmc_ffu_wait_busy(struct mmc_card *card) {
> > + int ret, busy = 0;
> > + struct mmc_command cmd = {0};
> > +
> > + memset(&cmd, 0, sizeof(struct mmc_command));
> > + cmd.opcode = MMC_SEND_STATUS;
> > + cmd.arg = card->rca << 16;
> > + cmd.flags = MMC_RSP_SPI_R2 | MMC_RSP_R1 | MMC_CMD_AC;
> > +
> > + do {
> > + ret = mmc_wait_for_cmd(card->host, &cmd, 0);
> > + if (ret)
> > + break;
> > +
> > + if (!busy && mmc_ffu_busy(&cmd)) {
> > + busy = 1;
> > + if (card->host->caps & MMC_CAP_WAIT_WHILE_BUSY) {
> > + pr_warn("%s: Warning: Host did not "
> > + "wait for busy state to end.\n",
> > + mmc_hostname(card->host));
> > + }
> > + }
> > +
> > + } while (mmc_ffu_busy(&cmd));
> > +
> > + return ret;
> > +}
>
> By just browsing through the code for preparing and handling the mmc data
> request in the above functions, I find quite some duplication of code from
> the card/*.c files. Could we maybe share some code between these layers to
> avoid duplications, and thus errors?
Yes, we are aware of some duplications and tried to avoid it.
Eventually, sharing code to avoid duplications will require changes in existing
card/*c files.
We preferred not changing existing code as much as possible.
>
> > +
> > +/*
> > + * transfer with certain parameters
> > + */
> > +static int mmc_ffu_simple_transfer(struct mmc_card *card,
> > + struct scatterlist *sg, unsigned int sg_len, u32 arg,
> > + unsigned int blocks, unsigned int blksz) {
> > + struct mmc_request mrq = {0};
> > + struct mmc_command cmd = {0};
> > + struct mmc_command stop = {0};
> > + struct mmc_data data = {0};
> > +
> > + mrq.cmd = &cmd;
> > + mrq.data = &data;
> > + mrq.stop = &stop;
> > + mmc_ffu_prepare_mrq(card, &mrq, sg, sg_len, arg, blocks, blksz);
> > + mmc_wait_for_req(card->host, &mrq);
> > +
> > + mmc_ffu_wait_busy(card);
> > +
> > + return mmc_ffu_check_result(&mrq); }
> > +
> > +/*
> > + * Map memory into a scatterlist.
> > + */
> > +static unsigned int mmc_ffu_map_sg(struct mmc_ffu_mem *mem, int
> size,
> > + struct scatterlist *sglist, unsigned int max_segs,
> > + unsigned int max_seg_sz)
> > +{
> > + struct scatterlist *sg = sglist;
> > + unsigned int i;
> > + unsigned long sz = size;
> > + unsigned int sctr_len = 0;
> > + unsigned long len;
> > +
> > + sg_init_table(sglist, max_segs);
> > +
> > + for (i = 0; i < mem->cnt && sz; i++, sz -= len) {
> > + len = PAGE_SIZE * (1 << mem->arr[i].order);
> > +
> > + if (len > sz) {
> > + len = sz;
> > + sz = 0;
> > + }
> > +
> > + sg_set_page(sg, mem->arr[i].page, len, 0);
> > + sg = sg_next(sg);
> > + sctr_len += 1;
> > + }
> > + sg_mark_end(sg);
> > +
> > + return sctr_len;
> > +}
> > +
> > +static void mmc_ffu_free_mem(struct mmc_ffu_mem *mem) {
> > + if (!mem)
> > + return;
> > +
> > + while (mem->cnt--)
> > + __free_pages(mem->arr[mem->cnt].page,
> > + mem->arr[mem->cnt].order);
> > +
> > + kfree(mem->arr);
> > + kfree(mem);
> > +}
> > +
> > +/*
> > + * Cleanup struct mmc_ffu_area.
> > + */
> > +static int mmc_ffu_area_cleanup(struct mmc_ffu_area *area) {
> > + kfree(area->sg);
> > + mmc_ffu_free_mem(area->mem);
> > +
> > + return 0;
> > +}
> > +
> > +/*
> > + * Allocate a lot of memory, preferably max_sz but at least min_sz.
> > +In case
> > + * there isn't much memory do not exceed 1/16th total low mem pages.
> > +Also do
> > + * not exceed a maximum number of segments and try not to make
> > +segments much
> > + * bigger than maximum segment size.
> > + */
> > +static struct mmc_ffu_mem *mmc_ffu_alloc_mem(unsigned long min_sz,
> > + unsigned long max_sz, unsigned int max_segs, unsigned int
> > +max_seg_sz) {
> > + unsigned long max_page_cnt = DIV_ROUND_UP(max_sz, PAGE_SIZE);
> > + unsigned long min_page_cnt = DIV_ROUND_UP(min_sz, PAGE_SIZE);
> > + unsigned long max_seg_page_cnt = DIV_ROUND_UP(max_seg_sz,
> PAGE_SIZE);
> > + unsigned long page_cnt = 0;
> > + /* we divide by 16 to ensure we will not allocate a big amount
> > + * of unnecessary pages */
> > + unsigned long limit = nr_free_buffer_pages() >> 4;
> > + struct mmc_ffu_mem *mem;
> > + gfp_t flags = GFP_KERNEL | GFP_DMA | __GFP_NOWARN |
> > +__GFP_NORETRY;
> > +
> > + if (max_page_cnt > limit)
> > + max_page_cnt = limit;
> > +
> > + if (min_page_cnt > max_page_cnt)
> > + min_page_cnt = max_page_cnt;
> > +
> > + if (max_segs * max_seg_page_cnt > max_page_cnt)
> > + max_segs = DIV_ROUND_UP(max_page_cnt,
> > + max_seg_page_cnt);
> > +
> > + mem = kzalloc(sizeof(struct mmc_ffu_mem), GFP_KERNEL);
> > + if (!mem)
> > + return NULL;
> > +
> > + mem->arr = kzalloc(sizeof(struct mmc_ffu_pages) * max_segs,
> > + GFP_KERNEL);
> > + if (!mem->arr)
> > + goto out_free;
> > +
> > + while (max_page_cnt) {
> > + struct page *page;
> > + unsigned int order;
> > +
> > + order = get_order(max_seg_page_cnt << PAGE_SHIFT);
> > +
> > + do {
> > + page = alloc_pages(flags, order);
> > + } while (!page && order--);
> > +
> > + if (!page)
> > + goto out_free;
> > +
> > + mem->arr[mem->cnt].page = page;
> > + mem->arr[mem->cnt].order = order;
> > + mem->cnt += 1;
> > + if (max_page_cnt <= (1UL << order))
> > + break;
> > + max_page_cnt -= 1UL << order;
> > + page_cnt += 1UL << order;
> > + }
> > +
> > + if (page_cnt < min_page_cnt)
> > + goto out_free;
> > +
> > + return mem;
> > +
> > +out_free:
> > + mmc_ffu_free_mem(mem);
> > + return NULL;
> > +}
> > +
> > +/*
> > + * Initialize an area for data transfers.
> > + * Copy the data to the allocated pages.
> > + */
> > +static int mmc_ffu_area_init(struct mmc_ffu_area *area, struct
> mmc_card *card,
> > + const u8 *data, int size)
> > +{
> > + int ret;
> > + int i;
> > + int length = 0;
> > +
> > + area->max_tfr = size;
> > +
> > + /* Try to allocate enough memory for a max. sized transfer. Less is
> > OK
> > + * because the same memory can be mapped into the scatterlist more
> than
> > + * once. Also, take into account the limits imposed on scatterlist
> > + * segments by the host driver.
> > + */
> > + area->mem = mmc_ffu_alloc_mem(1, area->max_tfr, area-
> >max_segs,
> > + area->max_seg_sz);
> > + if (!area->mem)
> > + return -ENOMEM;
> > +
> > + /* copy data to page */
> > + for (i = 0; i < area->mem->cnt; i++) {
> > + if (length > size) {
> > + ret = -EINVAL;
> > + goto out_free;
> > + }
> > +
> > + memcpy(page_address(area->mem->arr[i].page), data + length,
> > + min(size - length, (int)area->max_seg_sz));
> > + length += area->max_seg_sz;
> > + }
> > +
> > + area->sg = kmalloc(sizeof(struct scatterlist) * area->mem->cnt,
> > + GFP_KERNEL);
> > + if (!area->sg) {
> > + ret = -ENOMEM;
> > + goto out_free;
> > + }
> > +
> > + area->sg_len = mmc_ffu_map_sg(area->mem, size, area->sg,
> > + area->max_segs, area->mem->cnt);
> > +
> > + return 0;
> > +
> > +out_free:
> > + mmc_ffu_area_cleanup(area);
> > + return ret;
> > +}
> > +
> > +static int mmc_ffu_write(struct mmc_card *card, const u8 *src, u32 arg,
> > + int size)
> > +{
> > + int rc;
> > + struct mmc_ffu_area area;
> > + int max_tfr;
> > +
> > + area.sg = NULL;
> > + area.mem = NULL;
> > + area.max_segs = card->host->max_segs;
> > + area.max_seg_sz = card->host->max_seg_size & ~(CARD_BLOCK_SIZE
> - 1);
> > + do {
> > + max_tfr = size;
> > + if (max_tfr >> 9 > card->host->max_blk_count)
> > + max_tfr = card->host->max_blk_count << 9;
> > + if (max_tfr > card->host->max_req_size)
> > + max_tfr = card->host->max_req_size;
> > + if (DIV_ROUND_UP(max_tfr, area.max_seg_sz) > area.max_segs)
> > + max_tfr = area.max_segs * area.max_seg_sz;
> > +
> > + rc = mmc_ffu_area_init(&area, card, src, max_tfr);
> > + if (rc != 0)
> > + goto exit;
> > +
> > + rc = mmc_ffu_simple_transfer(card, area.sg, area.sg_len,
> > arg,
> > + max_tfr / CARD_BLOCK_SIZE, CARD_BLOCK_SIZE);
> > + if (rc != 0)
> > + goto exit;
> > +
> > + src += max_tfr;
> > + size -= max_tfr;
> > + } while (size > 0);
> > +
> > +exit:
> > + mmc_ffu_area_cleanup(&area);
> > + return rc;
> > +}
> > +
> > +/* Flush all scheduled work from the MMC work queue.
> > + * and initialize the MMC device */
> > +static int mmc_ffu_restart(struct mmc_card *card) {
> > + struct mmc_host *host = card->host;
> > + int err = 0;
> > +
> > + mmc_cache_ctrl(host, 0);
>
> Isn't cache flushing a bit late to consider here?
I think we need to call mmc_flush_cache() instead of mmc_cache_ctrl().
>
> I would, as stated earlier, prefer one FFU operational code. Thus it would be
> possible, before starting the FFU seuquence, to for example do cache
> flushing, stopping BKOPS and suspend/flush the blk queue etc.
> I think this would be more safe.
Agree, it probably would be more safe, but:
- as mentioned before, we would like to split the process into two operational
codes due
to different host power implementations
- We implemented FFU as non-blocking procedure, which allows host sending
regular R/W commands
between downloading FW data process and FW install
>
> > + err = mmc_power_save_host(host);
>
> This is fragile. There are no guarantees the eMMC card will be properly
> power cycled, I guess that is the requirement.
The FFU requirement is not necessarily making power cycle, but performing "eMMC
init"
sequence (CMD0, CMD1,...).
Once host cycling the power, it will run eMMC init sequence anyway.
>
> In some cases we can power cycle the card in some cases not, depending on
> the hardware design. Are there other ways of telling the card to start using
> the new firmware, but doing a full power cycle?
>
> Another concern I see, is if the card might change identity due to the
> firmware upgrade. The card identity is validated each system resume cycle
> and failed validation is an error we can't recover from.
The current assumption is that device will change its identity during the FFU.
But your concern is correct.
In order to support this, the mmc_init_card() function call should be modified
and assume that the card is not "oldcard".
>
> > + if (err) {
> > + pr_warn("%s: going to sleep failed (%d)!!!\n",
> > + __func__ , err);
> > + goto exit;
> > + }
> > +
> > + err = mmc_power_restore_host(host);
> > +
> > +exit:
> > +
> > + return err;
> > +}
> > +
> > +int mmc_ffu_download(struct mmc_card *card, const char *name) {
> > + u8 ext_csd[CARD_BLOCK_SIZE];
> > + int err;
> > + int ret;
> > + u32 arg;
> > + u32 fw_prog_bytes;
> > + const struct firmware *fw;
> > +
> > + /* Check if FFU is supported */
> > + if (!card->ext_csd.ffu_capable) {
> > + pr_err("FFU: %s: error FFU is not supported %d rev %d\n",
> > + mmc_hostname(card->host), card->ext_csd.ffu_capable,
> > + card->ext_csd.rev);
> > + return -EOPNOTSUPP;
> > + }
> > +
> > + if (strlen(name) > 512) {
> > + pr_err("FFU: %s: %.20s is not a valid argument\n",
> > + mmc_hostname(card->host), name);
> > + return -EINVAL;
> > + }
> > +
> > + /* setup FW data buffer */
> > + err = request_firmware(&fw, name, &card->dev);
> > + if (err) {
> > + pr_err("FFU: %s: Firmware request failed %d\n",
> > + mmc_hostname(card->host), err);
> > + return err;
> > + }
> > + if ((fw->size % CARD_BLOCK_SIZE)) {
> > + pr_warn("FFU: %s: Warning %zd firmware data size "
> > + "is not aligned!!!\n", mmc_hostname(card->host),
> > + fw->size);
> > + }
> > +
> > + mmc_get_card(card);
> > +
> > + /* set device to FFU mode */
> > + err = mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
> EXT_CSD_MODE_CONFIG,
> > + MMC_FFU_MODE_SET, card->ext_csd.generic_cmd6_time);
> > + if (err) {
> > + pr_err("FFU: %s: error %d FFU is not supported\n",
> > + mmc_hostname(card->host), err);
> > + goto exit;
> > + }
> > +
> > + /* Read the EXT_CSD */
> > + err = mmc_send_ext_csd(card, ext_csd);
> > + if (err) {
> > + pr_err("FFU: %s: error %d sending ext_csd\n",
> > + mmc_hostname(card->host), err);
> > + goto exit;
> > + }
> > +
> > + /* set CMD ARG */
> > + arg = ext_csd[EXT_CSD_FFU_ARG] |
> > + ext_csd[EXT_CSD_FFU_ARG + 1] << 8 |
> > + ext_csd[EXT_CSD_FFU_ARG + 2] << 16 |
> > + ext_csd[EXT_CSD_FFU_ARG + 3] << 24;
> > +
> > + err = mmc_ffu_write(card, fw->data, arg, (int)fw->size);
> > +
> > + /* host switch back to work in normal MMC Read/Write commands */
> > + ret = mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
> > + EXT_CSD_MODE_CONFIG, MMC_FFU_MODE_NORMAL,
> > + card->ext_csd.generic_cmd6_time);
> > + if (ret) {
> > + err = ret;
> > + goto exit;
> > + }
> > +
> > + /* Read the EXT_CSD */
> > + err = mmc_send_ext_csd(card, ext_csd);
> > + if (err) {
> > + pr_err("FFU: %s: error %d sending ext_csd\n",
> > + mmc_hostname(card->host), err);
> > + goto exit;
> > + }
> > +
> > + /* check that the eMMC has received the payload */
> > + fw_prog_bytes = ext_csd[EXT_CSD_NUM_OF_FW_SEC_PROG] |
> > + ext_csd[EXT_CSD_NUM_OF_FW_SEC_PROG + 1] << 8 |
> > + ext_csd[EXT_CSD_NUM_OF_FW_SEC_PROG + 2] << 16 |
> > + ext_csd[EXT_CSD_NUM_OF_FW_SEC_PROG + 3] << 24;
> > +
> > + /* convert sector to bytes */
> > + fw_prog_bytes *=
> > + CARD_BLOCK_SIZE << (ext_csd[EXT_CSD_DATA_SECTOR_SIZE] *
> 3);
> > + if (fw_prog_bytes != fw->size) {
> > + err = -EINVAL;
> > + pr_err("FFU: %s: error %d number of programmed fw sector\n",
> > + mmc_hostname(card->host), err);
> > + }
> > +
> > +exit:
> > + release_firmware(fw);
> > + mmc_put_card(card);
> > + return err;
> > +}
> > +EXPORT_SYMBOL(mmc_ffu_download);
> > +
> > +int mmc_ffu_install(struct mmc_card *card) {
> > + u8 ext_csd[CARD_BLOCK_SIZE];
> > + int err;
> > + u32 ffu_data_len;
> > + u32 timeout;
> > +
> > + /* Check if FFU is supported */
> > + if (!card->ext_csd.ffu_capable) {
> > + pr_err("FFU: %s: error FFU is not supported\n",
> > + mmc_hostname(card->host));
> > + return -EOPNOTSUPP;
> > + }
> > +
> > + err = mmc_send_ext_csd(card, ext_csd);
> > + if (err) {
> > + pr_err("FFU: %s: error %d sending ext_csd\n",
> > + mmc_hostname(card->host), err);
> > + return err;
> > + }
> > +
> > + /* check mode operation */
> > + if (!FFU_FEATURES(ext_csd[EXT_CSD_FFU_FEATURES])) {
> > + /* restart the eMMC */
> > + err = mmc_ffu_restart(card);
> > + if (err) {
> > + pr_err("FFU: %s: install error %d:\n",
> > + mmc_hostname(card->host), err);
> > + return err;
> > + }
> > + } else {
> > +
> > + ffu_data_len = ext_csd[EXT_CSD_NUM_OF_FW_SEC_PROG]|
> > + ext_csd[EXT_CSD_NUM_OF_FW_SEC_PROG + 1] << 8 |
> > + ext_csd[EXT_CSD_NUM_OF_FW_SEC_PROG + 2] << 16 |
> > + ext_csd[EXT_CSD_NUM_OF_FW_SEC_PROG + 3] << 24;
> > +
> > + if (!ffu_data_len) {
> > + err = -EPERM;
> > + return err;
> > + }
> > + /* set device to FFU mode */
> > + err = mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
> > + EXT_CSD_MODE_CONFIG, 0x1,
> > + card->ext_csd.generic_cmd6_time);
> > +
> > + if (err) {
> > + pr_err("FFU: %s: error %d FFU is not supported\n",
> > + mmc_hostname(card->host), err);
> > + return err;
> > + }
> > +
> > + timeout = ext_csd[EXT_CSD_OPERATION_CODE_TIMEOUT];
> > + if (timeout == 0 || timeout > 0x17) {
> > + timeout = 0x17;
> > + pr_warn("FFU: %s: operation code timeout is out "
> > + "of range. Using maximum timeout.\n",
> > + mmc_hostname(card->host));
> > + }
> > +
> > + /* timeout is at millisecond resolution */
> > + timeout = (100 * (1 << timeout) / 1000) + 1;
> > +
> > + /* set ext_csd to install mode */
> > + err = mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
> > + EXT_CSD_MODE_OPERATION_CODES,
> > + MMC_FFU_INSTALL_SET, timeout);
> > +
> > + if (err) {
> > + pr_err("FFU: %s: error %d setting install mode\n",
> > + mmc_hostname(card->host), err);
> > + return err;
> > + }
> > + }
> > +
> > + /* read ext_csd */
> > + err = mmc_send_ext_csd(card, ext_csd);
> > + if (err) {
> > + pr_err("FFU: %s: error %d sending ext_csd\n",
> > + mmc_hostname(card->host), err);
> > + return err;
> > + }
> > +
> > + /* return status */
> > + err = ext_csd[EXT_CSD_FFU_STATUS];
> > + if (err) {
> > + pr_err("FFU: %s: error %d FFU install:\n",
> > + mmc_hostname(card->host), err);
> > + return -EINVAL;
> > + }
> > +
> > + return 0;
> > +}
> > +EXPORT_SYMBOL(mmc_ffu_install);
> > diff --git a/drivers/mmc/core/mmc.c b/drivers/mmc/core/mmc.c index
> > 98e9eb0..a29065a 100644
> > --- a/drivers/mmc/core/mmc.c
> > +++ b/drivers/mmc/core/mmc.c
> > @@ -571,6 +571,14 @@ static int mmc_read_ext_csd(struct mmc_card
> *card, u8 *ext_csd)
> > card->ext_csd.data_sector_size = 512;
> > }
> >
> > + /* eMMC v5 or later */
> > + if (card->ext_csd.rev >= 7) {
> > + card->ext_csd.ffu_capable =
> > + ((ext_csd[EXT_CSD_SUPPORTED_MODE] & 1) == 1) &&
> > + ((ext_csd[EXT_CSD_FW_CONFIG] & 1) == 0);
> > + } else {
> > + card->ext_csd.ffu_capable = false;
> > + }
>
> I would like to add the "ffu_capable" field from a standalone separate patch.
No problem. Let us know if you prefer us doing this.
>
> > out:
> > return err;
> > }
> > diff --git a/include/linux/mmc/card.h b/include/linux/mmc/card.h index
> > b730272..bc6f6d0 100644
> > --- a/include/linux/mmc/card.h
> > +++ b/include/linux/mmc/card.h
> > @@ -87,6 +87,7 @@ struct mmc_ext_csd {
> > unsigned int data_tag_unit_size; /* DATA TAG UNIT
> > size */
> > unsigned int boot_ro_lock; /* ro lock support
> > */
> > bool boot_ro_lockable;
> > + bool ffu_capable; /* FFU support */
> > u8 raw_exception_status; /* 54 */
> > u8 raw_partition_support; /* 160 */
> > u8 raw_rpmb_size_mult; /* 168 */
> > diff --git a/include/linux/mmc/ffu.h b/include/linux/mmc/ffu.h new
> > file mode 100644 index 0000000..7e4133d
> > --- /dev/null
> > +++ b/include/linux/mmc/ffu.h
> > @@ -0,0 +1,51 @@
> > +/*
> > + *
> > + * ffu.h
> > + *
> > + * Copyright (c) 2013 SanDisk Corp.
> > + *
> > + * This program is free software; you can redistribute it and/or
> > +modify
> > + * it under the terms of the GNU General Public License as published
> > +by
> > + * the Free Software Foundation; either version 2 of the License, or
> > +(at
> > + * your option) any later version.
> > + *
> > + * This program was created by SanDisk Corp
> > + * The ffu.h file is obtained under the GPL v2.0 license that is
> > + * available via http://www.gnu.org/licenses/,
> > + * or http://www.opensource.org/licenses/gpl-2.0.php
> > +*/
> > +
> > +#if !defined(_FFU_H_)
> > +#define _FFU_H_
> > +
> > +#include <linux/mmc/card.h>
> > +
> > +#define CARD_BLOCK_SIZE 512
> > +
> > +/*
> > + * eMMC5.0 Field Firmware Update (FFU) opcodes */ #define
> > +MMC_FFU_DOWNLOAD_OP 302 #define MMC_FFU_INSTALL_OP 303
> > +
> > +#define MMC_FFU_MODE_SET 0x1
> > +#define MMC_FFU_MODE_NORMAL 0x0
> > +#define MMC_FFU_INSTALL_SET 0x1
> > +
> > +#ifdef CONFIG_MMC_FFU
> > +#define MMC_FFU_FEATURES 0x1
> > +#define FFU_FEATURES(ffu_features) (ffu_features &
> MMC_FFU_FEATURES)
> > +
> > +int mmc_ffu_download(struct mmc_card *card, const char *name); int
> > +mmc_ffu_install(struct mmc_card *card); #else static inline int
> > +mmc_ffu_download(struct mmc_card *card, const char *name) {
> > + return -ENOSYS;
> > +}
> > +static inline int mmc_ffu_install(struct mmc_card *card) {
> > + return -ENOSYS;
> > +}
> > +#endif
> > +#endif /* FFU_H_ */
> > diff --git a/include/linux/mmc/mmc.h b/include/linux/mmc/mmc.h index
> > 50bcde3..3651449 100644
> > --- a/include/linux/mmc/mmc.h
> > +++ b/include/linux/mmc/mmc.h
> > @@ -272,6 +272,9 @@ struct _mmc_csd {
> > * EXT_CSD fields
> > */
> >
> > +#define EXT_CSD_FFU_STATUS 26 /* R */
> > +#define EXT_CSD_MODE_OPERATION_CODES 29 /* W */
> > +#define EXT_CSD_MODE_CONFIG 30 /* R/W */
> > #define EXT_CSD_FLUSH_CACHE 32 /* W */
> > #define EXT_CSD_CACHE_CTRL 33 /* R/W */
> > #define EXT_CSD_POWER_OFF_NOTIFICATION 34 /* R/W */
> > @@ -290,6 +293,7 @@ struct _mmc_csd {
> > #define EXT_CSD_SANITIZE_START 165 /* W */
> > #define EXT_CSD_WR_REL_PARAM 166 /* RO */
> > #define EXT_CSD_RPMB_MULT 168 /* RO */
> > +#define EXT_CSD_FW_CONFIG 169 /* R/W */
> > #define EXT_CSD_BOOT_WP 173 /* R/W */
> > #define EXT_CSD_ERASE_GROUP_DEF 175 /* R/W */
> > #define EXT_CSD_PART_CONFIG 179 /* R/W */
> > @@ -325,6 +329,11 @@ struct _mmc_csd {
> > #define EXT_CSD_POWER_OFF_LONG_TIME 247 /* RO */
> > #define EXT_CSD_GENERIC_CMD6_TIME 248 /* RO */
> > #define EXT_CSD_CACHE_SIZE 249 /* RO, 4 bytes */
> > +#define EXT_CSD_NUM_OF_FW_SEC_PROG 302 /* RO, 4 bytes */
> > +#define EXT_CSD_FFU_ARG 487 /* RO, 4 bytes */
> > +#define EXT_CSD_OPERATION_CODE_TIMEOUT 491 /* RO */
> > +#define EXT_CSD_FFU_FEATURES 492 /* RO */
> > +#define EXT_CSD_SUPPORTED_MODE 493 /* RO */
> > #define EXT_CSD_TAG_UNIT_SIZE 498 /* RO */
> > #define EXT_CSD_DATA_TAG_SUPPORT 499 /* RO */
> > #define EXT_CSD_MAX_PACKED_WRITES 500 /* RO */
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-mmc"
> > in the body of a message to majord...@vger.kernel.org More
> majordomo
> > info at http://vger.kernel.org/majordomo-info.html
> >
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-mmc"
> > in the body of a message to majord...@vger.kernel.org More
> majordomo
> > info at http://vger.kernel.org/majordomo-info.html
>
> Kind regards
> Uffe
N�����r��y����b�X��ǧv�^�){.n�+����{��g"��^n�r���z���h�����&�� �G���h�(�階�ݢj"���m������z�ޖ���f���h���~�m�
