From: Bean Huo <[email protected]>
We can use GEN_CMD (CMD56) to read a data block with vendor-specific
format and meaning, which is very helpful and useful for customers to
analyze device health. This patch is to add support for this command.
At the same time, in order to make CMD56 universal, let more users of
mmc-utils can benefit from this. We will allow the user to enter [arg]
for special commands. Because some eMMC vendors provide an additional
set of commands beyond the GEN_CMD (CMD56) definition in the JEDEC
specification. These additional commands are implemented using generic
CMD56 commands, and they return a wealth of useful information about the
state of the NAND device. For these special purposes, the [31:1]stuff
bits in the CMD56 argument are used. You can refer to Micron eMMC
"TN-FC-32: e·MMC Device Health Report" as an example.
I didn't add data parsing, just print the raw data as it is vendor-specific.
Signed-off-by: Bean Huo <[email protected]>
Acked-by: Avri Altman <[email protected]>
Tested-by: Rossler Jakob (Nokia - DE/Ulm) <[email protected]>
---
V2:
1. Fix coding type issue
2. Add CMD56 argument default value 0x01 support, make [arg] optional
---
mmc.c | 8 ++++++++
mmc.h | 2 ++
mmc_cmds.c | 57 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
mmc_cmds.h | 1 +
4 files changed, 68 insertions(+)
diff --git a/mmc.c b/mmc.c
index eb2638b12271..6c563879cf1a 100644
--- a/mmc.c
+++ b/mmc.c
@@ -237,6 +237,14 @@ static struct Command commands[] = {
"secure-trim1 | secure-trim2 | trim \n",
NULL
},
+ { do_general_cmd_read, -1,
+ "gen_cmd read", "<device> [arg]\n"
+ "Send GEN_CMD (CMD56) to read vendor-specific format/meaning data from <device>\n\n"
+ "NOTE!: [arg] is optional and defaults to 0x1. If [arg] is specified, then [arg]\n"
+ "must be a 32-bit hexadecimal number, prefixed with 0x/0X. And bit0 in [arg] must\n"
+ "be 1.",
+ NULL
+ },
{ 0, 0, 0, 0 }
};
diff --git a/mmc.h b/mmc.h
index 25d6864ac76f..b621374a1ed1 100644
--- a/mmc.h
+++ b/mmc.h
@@ -41,6 +41,8 @@
[1] Discard Enable
[0] Identify Write Blocks for
Erase (or TRIM Enable) R1b */
+#define MMC_GEN_CMD 56 /* adtc [31:1] stuff bits.
+ [0]: RD/WR1 R1 */
#define R1_OUT_OF_RANGE (1 << 31) /* er, c */
#define R1_ADDRESS_ERROR (1 << 30) /* erx, c */
diff --git a/mmc_cmds.c b/mmc_cmds.c
index 49d3e324d266..bb0f0226bf23 100644
--- a/mmc_cmds.c
+++ b/mmc_cmds.c
@@ -2981,3 +2981,60 @@ out:
return ret;
#endif
}
+
+int do_general_cmd_read(int nargs, char **argv)
+{
+ int dev_fd;
+ char *device;
+ char *endptr;
+ __u8 buf[512];
+ __u32 arg = 0x01;
+ int ret = -EINVAL, i;
+ struct mmc_ioc_cmd idata;
+
+ if (nargs != 2 && nargs != 3) {
+ fprintf(stderr, "Usage: gen_cmd read </path/to/mmcblkX> [arg]\n");
+ exit(1);
+ }
+
+ device = argv[1];
+ dev_fd = open(device, O_RDWR);
+ if (dev_fd < 0) {
+ perror("device open failed");
+ exit(1);
+ }
+
+ /* arg is specified */
+ if (nargs == 3) {
+ arg = strtol(argv[2], &endptr, 16);
+ if (errno != 0 || *endptr != '\0' || !(arg & 0x1)) {
+ fprintf(stderr, "Wrong ARG, it should be Hex number and bit0 must be 1\n");
+ goto out;
+ }
+ }
+
+ memset(&idata, 0, sizeof(idata));
+ idata.write_flag = 0;
+ idata.opcode = MMC_GEN_CMD;
+ idata.arg = arg;
+ idata.flags = MMC_RSP_SPI_R1 | MMC_RSP_R1 | MMC_CMD_ADTC;
+ idata.blksz = 512;
+ idata.blocks = 1;
+ mmc_ioc_cmd_set_data(idata, buf);
+
+ ret = ioctl(dev_fd, MMC_IOC_CMD, &idata);
+ if (ret) {
+ perror("ioctl");
+ goto out;
+ }
+
+ printf("Data:\n");
+ for (i = 0; i < 512; i++) {
+ printf("%2x ", buf[i]);
+ if ((i + 1) % 16 == 0)
+ printf("\n");
+ }
+out:
+ close(dev_fd);
+ return ret;
+}
diff --git a/mmc_cmds.h b/mmc_cmds.h
index 8331ab2373fd..0f7c0041f753 100644
--- a/mmc_cmds.h
+++ b/mmc_cmds.h
@@ -46,3 +46,4 @@ int do_read_scr(int argc, char **argv);
int do_read_cid(int argc, char **argv);
int do_read_csd(int argc, char **argv);
int do_erase(int nargs, char **argv);
+int do_general_cmd_read(int nargs, char **argv);
--
2.34.1
On Sun, 2022-04-17 at 13:16 +0200, Bean Huo wrote:
> From: Bean Huo <[email protected]>
>
> We can use GEN_CMD (CMD56) to read a data block with vendor-specific
> format and meaning, which is very helpful and useful for customers to
> analyze device health. This patch is to add support for this command.
>
> At the same time, in order to make CMD56 universal, let more users of
> mmc-utils can benefit from this. We will allow the user to enter
> [arg]
> for special commands. Because some eMMC vendors provide an additional
> set of commands beyond the GEN_CMD (CMD56) definition in the JEDEC
> specification. These additional commands are implemented using
> generic
> CMD56 commands, and they return a wealth of useful information about
> the
> state of the NAND device. For these special purposes, the [31:1]stuff
> bits in the CMD56 argument are used. You can refer to Micron eMMC
> "TN-FC-32: e·MMC Device Health Report" as an example.
>
> I didn't add data parsing, just print the raw data as it is vendor-
> specific.
>
> Signed-off-by: Bean Huo <[email protected]>
> Acked-by: Avri Altman <[email protected]>
> Tested-by: Rossler Jakob (Nokia - DE/Ulm) <[email protected]>
> ---
Hi Ullf,
Would you be willing to merge this patch into your mmc-utils git repo?
Kind regards,
Bean
On Fri, 22 Apr 2022 at 21:41, Bean Huo <[email protected]> wrote:
>
> On Sun, 2022-04-17 at 13:16 +0200, Bean Huo wrote:
> > From: Bean Huo <[email protected]>
> >
> > We can use GEN_CMD (CMD56) to read a data block with vendor-specific
> > format and meaning, which is very helpful and useful for customers to
> > analyze device health. This patch is to add support for this command.
> >
> > At the same time, in order to make CMD56 universal, let more users of
> > mmc-utils can benefit from this. We will allow the user to enter
> > [arg]
> > for special commands. Because some eMMC vendors provide an additional
> > set of commands beyond the GEN_CMD (CMD56) definition in the JEDEC
> > specification. These additional commands are implemented using
> > generic
> > CMD56 commands, and they return a wealth of useful information about
> > the
> > state of the NAND device. For these special purposes, the [31:1]stuff
> > bits in the CMD56 argument are used. You can refer to Micron eMMC
> > "TN-FC-32: e·MMC Device Health Report" as an example.
> >
> > I didn't add data parsing, just print the raw data as it is vendor-
> > specific.
> >
> > Signed-off-by: Bean Huo <[email protected]>
> > Acked-by: Avri Altman <[email protected]>
> > Tested-by: Rossler Jakob (Nokia - DE/Ulm) <[email protected]>
> > ---
>
> Hi Ullf,
>
> Would you be willing to merge this patch into your mmc-utils git repo?
Certainly! My apologies for the delay!
Applied to mmc-utils.git master, thanks!
Kind regards
Uffe