if mmc_blk_alloc_part() fails for 'if (card->ext_csd.pt1_size)', why
would you want to give this code the opportunity to call
mmc_blk_alloc_part() again with the next if() below? If
mmc_blk_alloc_part() fails, is it a big problem, like kzalloc() failing?
If so, I would think you would want to immediately quit this function
and report an error (either use errno value or pass the value of ret).
--
J (James/Jay) Freyensee
Storage Technology Group
Intel Corporation
Please document that in your code. Please add a function header to this
function.
One or the other.
but a function comment header should be added at some point. Yes, the
mmc subsystem is pretty legacy, but any new work that gets added to the
kernel is usually has higher standards for stuff like not just adding
function header comments but putting them in a certain format...at least
this is what I went through when I had things accepted upstream.
I scanned over your changes to enable GP partitions, and while I
did not yet test your changes, they do appear to be ok.
Please CC me on the next version of your patch so I can give
it a spin and Ack on it.
I do suggest at least thinking about ways to improve on this. Back
when I added boot partition support, I already didn't like the fact
that the block driver had to be aware of the eMMC partitions, with all
the code duplication and such. By adding four more partitions,
you've compounded the problem and now you have six pieces of
code that really look the same.
The core/mmc.c right now scans the size values and stores them
in mmc_card, and the block driver is smart enough to know what
to do with them. Why not make the block code generic at the expense
of keeping an array of structures in mmc_card? The block driver
would then just iterate over these and create the additional devices. The partition
switch routine would then operate on these structures, and could then be pulled out
of the block driver, making it simpler.
Each physical partition could be described by something like -
struct mmc_part {
unsigned int size;
unsigned int cookie; /* for mmc, idx used in mmc_switch, part_type from current mmc_blk_data */
char name[DISK_NAME_LEN];
bool force_ro; /* to make boot parts RO by default */
};
Just an idea. Clearly, if MMC grows by another 4-5 possible partitions, it's not
realistic to be copy-pasting the same block of code, nor would it make sense
if/when these types of devices support an arbitrary amount of physical partitions.
Thanks,
A
Okay, I understand what you want. I will try to make patch included
your suggestion again.
Thanks for your review.
> It allows gerneral purpose partitions in MMC Device.
> And I try to simpliy make mmc_blk_alloc_parts using mmc_part structure suggested by Andrei Warkentin.
> After patching, we can see general purpose partitions like this.
>> cat /proc/partitions
> 179 0 847872 mmcblk0
> 179 192 4096 mmcblk0gp4
> 179 160 4096 mmcblk0gp3
> 179 128 4096 mmcblk0gp2
> 179 96 1052672 mmcblk0gp1
We notice there are 32 minors between 128 and 96,
how are these allocated?
> 179 64 1024 mmcblk0boot1
> 179 32 1024 mmcblk0boot0
Does this mean that each GP-partition can contain an MBR and thus several
subpartitions so we get things like:
mmcblk0gp1p1, mmcblk0gp1p2...
?
Yours,
Linus Walleij
Your review will always inspire me.
I will post again.
plz review after..
Thanks.
> A
>
Signed-off-by: Namjae Jeon <linki...@gmail.com>
---
drivers/mmc/card/block.c | 31 ++++++++++++++------------
drivers/mmc/core/mmc.c | 53 ++++++++++++++++++++++++++++++++++++++++++---
include/linux/mmc/card.h | 34 ++++++++++++++++++++++++++++-
include/linux/mmc/mmc.h | 5 +++-
4 files changed, 103 insertions(+), 20 deletions(-)
diff --git a/drivers/mmc/card/block.c b/drivers/mmc/card/block.c
index 1ff5486..8452872 100644
--- a/drivers/mmc/card/block.c
+++ b/drivers/mmc/card/block.c
@@ -1377,26 +1377,29 @@ static int mmc_blk_alloc_part(struct mmc_card *card,
return 0;
}
+/* MMC Physical partition consist of two boot partitions and
+ * up to four general purpose partitions.
+ * For each partitions enabled in EXT_CSD
+ * The block device will allocated to provide access to the partition.
+ */
+
static int mmc_blk_alloc_parts(struct mmc_card *card, struct mmc_blk_data *md)
{
- int ret = 0;
+ int idx, ret = 0;
if (!mmc_card_mmc(card))
return 0;
- if (card->ext_csd.boot_size) {
- ret = mmc_blk_alloc_part(card, md, EXT_CSD_PART_CONFIG_ACC_BOOT0,
- card->ext_csd.boot_size >> 9,
- true,
- "boot0");
- if (ret)
- return ret;
- ret = mmc_blk_alloc_part(card, md, EXT_CSD_PART_CONFIG_ACC_BOOT1,
- card->ext_csd.boot_size >> 9,
- true,
- "boot1");
- if (ret)
- return ret;
+ for (idx = 0; idx < card->nr_parts; idx++) {
+ if (card->part[idx].size) {
+ ret = mmc_blk_alloc_part(card, md,
+ card->part[idx].part_cfg,
+ card->part[idx].size >> 9,
+ card->part[idx].force_ro,
+ card->part[idx].name);
+ if (ret)
+ return ret;
+ }
}
return ret;
diff --git a/drivers/mmc/core/mmc.c b/drivers/mmc/core/mmc.c
index 5700b1c..f38bb6a 100644
--- a/drivers/mmc/core/mmc.c
+++ b/drivers/mmc/core/mmc.c
@@ -239,7 +239,9 @@ static int mmc_get_ext_csd(struct mmc_card *card, u8 **new_ext_csd)
*/
static int mmc_read_ext_csd(struct mmc_card *card, u8 *ext_csd)
{
- int err = 0;
+ int err = 0, idx;
+ unsigned int part_size;
+ u8 hc_erase_grp_sz = -EINVAL, hc_wp_grp_sz = -EINVAL;
BUG_ON(!card);
@@ -340,7 +342,15 @@ static int mmc_read_ext_csd(struct mmc_card *card, u8 *ext_csd)
* There are two boot regions of equal size, defined in
* multiples of 128K.
*/
- card->ext_csd.boot_size = ext_csd[EXT_CSD_BOOT_MULT] << 17;
+ if (ext_csd[EXT_CSD_BOOT_MULT]) {
+ for (idx = 0; idx < MMC_NUM_BOOT_PARTITION;
+ idx++) {
+ part_size = ext_csd[EXT_CSD_BOOT_MULT] << 17;
+ mmc_part_add(card, part_size,
+ EXT_CSD_PART_CONFIG_ACC_BOOT0 + idx,
+ "boot%d", idx, true);
+ }
+ }
}
card->ext_csd.raw_hc_erase_gap_size =
@@ -361,9 +371,9 @@ static int mmc_read_ext_csd(struct mmc_card *card, u8 *ext_csd)
*/
if ((ext_csd[EXT_CSD_PARTITION_SUPPORT] & 0x2) &&
(ext_csd[EXT_CSD_PARTITION_ATTRIBUTE] & 0x1)) {
- u8 hc_erase_grp_sz =
+ hc_erase_grp_sz =
ext_csd[EXT_CSD_HC_ERASE_GRP_SIZE];
- u8 hc_wp_grp_sz =
+ hc_wp_grp_sz =
ext_csd[EXT_CSD_HC_WP_GRP_SIZE];
card->ext_csd.enhanced_area_en = 1;
@@ -392,6 +402,41 @@ static int mmc_read_ext_csd(struct mmc_card *card, u8 *ext_csd)
card->ext_csd.enhanced_area_offset = -EINVAL;
card->ext_csd.enhanced_area_size = -EINVAL;
}
+
+ /*
+ * General purpose partition feature support --
+ * If ext_csd have the size of general purpose partitions,
+ * set size, part_cfg, partition name in mmc_part.
+ */
+
+ if (ext_csd[EXT_CSD_PARTITION_SUPPORT] &
+ EXT_CSD_PART_SUPPORT_PART_EN) {
+ if (hc_erase_grp_sz == -EINVAL ||
+ hc_wp_grp_sz == -EINVAL) {
+ hc_erase_grp_sz =
+ ext_csd[EXT_CSD_HC_ERASE_GRP_SIZE];
+ hc_wp_grp_sz =
+ ext_csd[EXT_CSD_HC_WP_GRP_SIZE];
+
+ card->ext_csd.enhanced_area_en = 1;
+ }
+
+ for (idx = 0; idx < MMC_NUM_GP_PARTITION; idx++) {
+ if (!ext_csd[EXT_CSD_GP_SIZE_MULT_X_Y] &&
+ !ext_csd[EXT_CSD_GP_SIZE_MULT_X_Y + 1] &&
+ !ext_csd[EXT_CSD_GP_SIZE_MULT_X_Y + 2])
+ continue;
+ part_size =
+ (ext_csd[EXT_CSD_GP_SIZE_MULT_X_Y + 2] << 16) +
+ (ext_csd[EXT_CSD_GP_SIZE_MULT_X_Y + 1] << 8) +
+ ext_csd[EXT_CSD_GP_SIZE_MULT_X_Y];
+ part_size *= (size_t)(hc_erase_grp_sz *
+ hc_wp_grp_sz);
+ mmc_part_add(card, part_size << 19,
+ EXT_CSD_PART_CONFIG_ACC_GP0 + idx,
+ "gp%d", idx, false);
+ }
+ }
card->ext_csd.sec_trim_mult =
ext_csd[EXT_CSD_SEC_TRIM_MULT];
card->ext_csd.sec_erase_mult =
diff --git a/include/linux/mmc/card.h b/include/linux/mmc/card.h
index b460fc2..4bff3d0 100644
--- a/include/linux/mmc/card.h
+++ b/include/linux/mmc/card.h
@@ -12,6 +12,7 @@
#include <linux/mmc/core.h>
#include <linux/mod_devicetable.h>
+#include <linux/genhd.h>
struct mmc_cid {
unsigned int manfid;
@@ -63,7 +64,6 @@ struct mmc_ext_csd {
bool enhanced_area_en; /* enable bit */
unsigned long long enhanced_area_offset; /* Units: Byte */
unsigned int enhanced_area_size; /* Units: KB */
- unsigned int boot_size; /* in bytes */
u8 raw_partition_support; /* 160 */
u8 raw_erased_mem_count; /* 181 */
u8 raw_ext_csd_structure; /* 194 */
@@ -157,6 +157,23 @@ struct sdio_func_tuple;
#define SDIO_MAX_FUNCS 7
+/* The number of MMC physical partitions
+ * It consist of boot partitions(2), general purpose partitions(4) in MMC v4.4
+ */
+#define MMC_NUM_BOOT_PARTITION 2
+#define MMC_NUM_GP_PARTITION 4
+#define MMC_NUM_PHY_PARTITION 6
+
+/*
+ * MMC Physical partitions
+ */
+struct mmc_part {
+ unsigned int size; /* partition size (in bytes) */
+ unsigned int part_cfg; /* it used to part_type */
+ char name[DISK_NAME_LEN];
+ bool force_ro; /* to make boot parts RO by default */
+};
+
/*
* MMC device
*/
@@ -216,9 +233,24 @@ struct mmc_card {
unsigned int sd_bus_speed; /* Bus Speed Mode set for the card */
struct dentry *debugfs_root;
+ struct mmc_part part[MMC_NUM_PHY_PARTITION]; /* physical partitions */
+ unsigned int nr_parts;
};
/*
+ * This function fill contents in mmc_part.
+ */
+static inline void mmc_part_add(struct mmc_card *card, unsigned int size,
+ unsigned int part_cfg, char *name, int idx, bool ro)
+{
+ card->part[card->nr_parts].size = size;
+ card->part[card->nr_parts].part_cfg = part_cfg;
+ sprintf(card->part[card->nr_parts].name, name, idx);
+ card->part[card->nr_parts].force_ro = ro;
+ card->nr_parts++;
+}
+
+/*
* The world is not perfect and supplies us with broken mmc/sdio devices.
* For at least some of these bugs we need a work-around.
*/
diff --git a/include/linux/mmc/mmc.h b/include/linux/mmc/mmc.h
index 5a794cb..284a1b0 100644
--- a/include/linux/mmc/mmc.h
+++ b/include/linux/mmc/mmc.h
@@ -270,6 +270,7 @@ struct _mmc_csd {
* EXT_CSD fields
*/
+#define EXT_CSD_GP_SIZE_MULT_X_Y 143 /* R/W */
#define EXT_CSD_PARTITION_ATTRIBUTE 156 /* R/W */
#define EXT_CSD_PARTITION_SUPPORT 160 /* RO */
#define EXT_CSD_WR_REL_PARAM 166 /* RO */
@@ -302,7 +303,9 @@ struct _mmc_csd {
#define EXT_CSD_PART_CONFIG_ACC_MASK (0x7)
#define EXT_CSD_PART_CONFIG_ACC_BOOT0 (0x1)
-#define EXT_CSD_PART_CONFIG_ACC_BOOT1 (0x2)
+#define EXT_CSD_PART_CONFIG_ACC_GP0 (0x4)
+
+#define EXT_CSD_PART_SUPPORT_PART_EN (0x1)
#define EXT_CSD_CMD_SET_NORMAL (1<<0)
#define EXT_CSD_CMD_SET_SECURE (1<<1)
--
1.7.4.4
--
I receive mmc-next you inform like this.
git clone git://dev.laptop.org/users/cjb/mmc
What I did wrong ?
Thanks.
On Tue, Oct 04 2011, NamJae Jeon wrote:
>> Still doesn't apply here. Are you sure you're on the "mmc-next" branch
>> of the repository?
>
> I receive mmc-next you inform like this.
>
> git clone git://dev.laptop.org/users/cjb/mmc
>
> What I did wrong ?
You need to type "git checkout mmc-next" afterwards to switch to the
mmc-next branch.
Thanks,
No. 0,1 are an accidental result of the order in which ext-csd is
parsed.
> partition is variable in 2,3,4,5,6 of array.
> If How someone change code, problem occur ?
If you move code around in mmc_read_ext_csd() the order can change,
hence it is fragile.
----- Original Message -----
> From: "Namjae Jeon" <linki...@gmail.com>
> To: c...@laptop.org, linu...@vger.kernel.org
> Cc: linux-...@vger.kernel.org, awark...@vmware.com, "adrian hunter" <adrian...@intel.com>, "linus walleij"
> <linus....@linaro.org>, "james p freyensee" <james_p_...@linux.intel.com>, seb...@gmail.com, "Ulf Hansson"
> <Ulf.H...@stericsson.com>, "stefan xk nilsson" <stefan.x...@stericsson.com>, "per forlin"
> <per.f...@stericsson.com>, "johan rudholm" <johan....@stericsson.com>, "Namjae Jeon" <linki...@gmail.com>
> + for (idx = 0; idx < card->nr_parts; idx++) {
> + if (mmc_is_boot_partition(&card->part[idx]) &&
> + !mmc_boot_partition_access(card->host))
> + continue;
> + if (card->part[idx].size) {
> + ret = mmc_blk_alloc_part(card, md,
> + card->part[idx].part_cfg,
> + card->part[idx].size >> 9,
> + card->part[idx].force_ro,
> + card->part[idx].name);
> + if (ret)
> + return ret;
> + }
> }
Is there any reason for putting the mmc_boot_partition_access() logic here? If
boot partitions are not allowed by host, then just don't add the mmc_parts to
the parts array, no? Such minutae should belong in core mmc code, not block driver, IMHO.
A
2011/10/5 NamJae Jeon <linki...@gmail.com>:
>
> How about changing patch like this ?
>
> if (ext_csd[EXT_CSD_BOOT_MULT] && mmc_boot_partition_access(card->host)) {
> for (idx = 0; idx < MMC_NUM_BOOT_PARTITION; idx++) {
> part_size = ext_csd[EXT_CSD_BOOT_MULT] << 17;
> mmc_part_add(card, part_size,
> EXT_CSD_PART_CONFIG_ACC_BOOT0 + idx,
> "boot%d", idx, true);
> }
> }
>
> I want to hear the opinion from andrei and
> adrian(mmc_boot_partition_access patch's author)
> Thanks.
I like this.
A
Acked-by: Andrei Warkentin <andrey.w...@gmail.com>
A