Hey Christo,
On 10-05-16 21:17, Christo Radev wrote:
Hi Olliver,
On Tuesday, May 10, 2016 at 9:46:45 PM UTC+3, Olliver Schinagl wrote:
Hey Christo,
On 10-05-16 19:25, Christo Radev wrote:
Hi Olliver,
I try to build Armbian with kernel 4.5.0-rc6 by adding the patch
as discussed here
<https://groups.google.com/d/msg/linux-sunxi/vSC4vx8_ok8/K0jW8bmcAQAJ>.
In addition I have to add following as well:
|
diff --git a/arch/arm/boot/dts/Makefileb/arch/arm/boot/dts/Makefile
index 95c1923..9d6cfa8100644
---a/arch/arm/boot/dts/Makefile
+++b/arch/arm/boot/dts/Makefile
@@-695,6+696,7@@dtb-$(CONFIG_MACH_SUN7I)+=\
sun7i-a20-olimex-som-evb.dtb \
sun7i-a20-olinuxino-lime.dtb \
sun7i-a20-olinuxino-lime2.dtb \
+ sun7i-a20-olinuxino-lime2-emmc.dtb \
sun7i-a20-olinuxino-micro.dtb \
sun7i-a20-orangepi.dtb \
sun7i-a20-orangepi-mini.dtb \
|
to be able to compile the new sun7i-a20-olinuxino-lime2-emmc.dts
file.
Good point, I actually forgot that bit in my own Makefile.
After booting I have renamed it to sun7i-a20-olinuxino-lime2.dts
and reboot lime-eMMC board.
why rename it to dts? i'm sure you mean dtb, and I'm guessing
because armbian loads that per default.
Yes, currently I use LIME2 configuration and this way is faster. Later
on I will make new configuration set for Lime2-eMMC.
Unfortunately, I get some error messages:
|
root@lime2:~# dmesg | grep mmc
[0.000000]Kernelcommand line:console=tty1 root=/dev/mmcblk0p1
rootwait rootfstype=ext4 cgroup_enable=memory
swapaccount=1sunxi_ve_mem_reserve=0sunxi_g2d_mem_reserve=0sunxi_no_mali_mem_reserve
sunxi_fb_mem_reserve=16hdmi.audio=EDID:0disp.screen0_output_mode=1920x1080p60panic=10consoleblank=0enforcing=0loglevel=1
[3.721024]sunxi-mmc 1c0f000.mmc:GotCD GPIO
[3.758926]sunxi-mmc 1c0f000.mmc:base:0xf08dc000irq:26
[3.759832]sunxi-mmc 1c11000.mmc:allocated mmc-pwrseq
[4.528643]sunxi-mmc 1c11000.mmc:fatal err update clk timeout
[4.535700]mmc0:host does notsupport reading read-only
switch,assuming write-enable
[4.538253]mmc0:newhigh speed SDHC card at address 0002
[4.539066]mmcblk0:mmc0:0002000003.70GiB
[4.540959] mmcblk0:p1
[4.548773]sunxi-mmc 1c11000.mmc:base:0xf08f2000irq:27
[6.418686]sunxi-mmc 1c11000.mmc:fatal err update clk timeout
[6.600359]EXT4-fs (mmcblk0p1):mounted filesystem withwriteback
data mode.Opts:(null)
[7.168650]sunxi-mmc 1c11000.mmc:fatal err update clk timeout
[7.918646]sunxi-mmc 1c11000.mmc:fatal err update clk timeout
[9.875182]EXT4-fs
(mmcblk0p1):re-mounted.Opts:commit=600,errors=remount-ro
|
Where could be the problem?
Not exactly sure, are you using the correct lime ;)
Of course, I use A20-Olinuxino-Lome2-eMMC boards for testing.
It does look like the correct dtb was used, since the second mmc
controller is available. Maybe the bootloader prepares/inits
something that is missing? sounds unlikly but possible. Maybe
armbian does more changes?
For sure in the beginning I have patched and enabled eMMC in the
kernel only (without any changes in u-boot). And it has worked fine.
Later on I have patched u-boot to recognize eMMC as second MMC device.
Meanwhile, I have build Armbian dev image with u-boot v2016.05-rc3 and
kernel 4.6-rc7 using may patch for eMMC:
|
diff --git a/arch/arm/boot/dts/sun7i-a20-olinuxino-lime2.dts
b/arch/arm/boot/dts/sun7i-a20-olinuxino-lime2.dts
index d5c796c..1f5339d100644
---a/arch/arm/boot/dts/sun7i-a20-olinuxino-lime2.dts
+++b/arch/arm/boot/dts/sun7i-a20-olinuxino-lime2.dts
@@-188,6+188,15@@
status ="okay";
};
+&mmc2 {
+ pinctrl-names ="default";
+ pinctrl-0=<&mmc2_pins_a>;
+ vmmc-supply =<®_vcc3v3>;
+ bus-width =<8>;
+ non-removable;
+ status ="okay";
+};
+
&ohci0 {
status ="okay";
};
diff --git a/drivers/mmc/core/mmc.c b/drivers/mmc/core/mmc.c
index 3d5087b..78668aa100644
---a/drivers/mmc/core/mmc.c
+++b/drivers/mmc/core/mmc.c
@@-504,7+504,7@@staticintmmc_decode_ext_csd(structmmc_card *card,u8
*ext_csd)
pr_info("%s: MAN_BKOPS_EN bit is not set\n",
mmc_hostname(card->host));
}
-
+#if 0
|
|you should remove this ...|
|
/* check whether the eMMC card supports HPI */
if(!broken_hpi &&(ext_csd[EXT_CSD_HPI_FEATURES]&0x1)){
card->ext_csd.hpi =1;
@@-519,7+519,7@@staticintmmc_decode_ext_csd(structmmc_card *card,u8
*ext_csd)
card->ext_csd.out_of_int_time =
ext_csd[EXT_CSD_OUT_OF_INTERRUPT_TIME]*10;
}
-
+#endif
|
|and this from your own patch-set.
Instead, the broken-hpi flag should be set in the mcc section, from my
patch:
emmc: emmc@0 {
reg = <0>;
compatible = "mmc-card";
broken-hpi;
};
|
|
card->ext_csd.rel_param =ext_csd[EXT_CSD_WR_REL_PARAM];
card->ext_csd.rst_n_function =ext_csd[EXT_CSD_RST_N_FUNCTION];
diff --git a/drivers/pinctrl/sunxi/pinctrl-sun7i-a20.c
b/drivers/pinctrl/sunxi/pinctrl-sun7i-a20.c
index cf1ce0c..9fc12d2100644
---a/drivers/pinctrl/sunxi/pinctrl-sun7i-a20.c
+++b/drivers/pinctrl/sunxi/pinctrl-sun7i-a20.c
@@-314,19+314,23@@staticconststructsunxi_desc_pin sun7i_a20_pins[]={
SUNXI_PIN(SUNXI_PINCTRL_PIN(C,12),
SUNXI_FUNCTION(0x0,"gpio_in"),
SUNXI_FUNCTION(0x1,"gpio_out"),
- SUNXI_FUNCTION(0x2,"nand0")),/* NDQ4 */
+ SUNXI_FUNCTION(0x2,"nand0"),/* NDQ4 */
+ SUNXI_FUNCTION(0x3,"mmc2")),/* D4 */
SUNXI_PIN(SUNXI_PINCTRL_PIN(C,13),
SUNXI_FUNCTION(0x0,"gpio_in"),
SUNXI_FUNCTION(0x1,"gpio_out"),
- SUNXI_FUNCTION(0x2,"nand0")),/* NDQ5 */
+ SUNXI_FUNCTION(0x2,"nand0"),/* NDQ5 */
+ SUNXI_FUNCTION(0x3,"mmc2")),/* D5 */
SUNXI_PIN(SUNXI_PINCTRL_PIN(C,14),
SUNXI_FUNCTION(0x0,"gpio_in"),
SUNXI_FUNCTION(0x1,"gpio_out"),
- SUNXI_FUNCTION(0x2,"nand0")),/* NDQ6 */
+ SUNXI_FUNCTION(0x2,"nand0"),/* NDQ6 */
+ SUNXI_FUNCTION(0x3,"mmc2")),/* D6 */
SUNXI_PIN(SUNXI_PINCTRL_PIN(C,15),
SUNXI_FUNCTION(0x0,"gpio_in"),
SUNXI_FUNCTION(0x1,"gpio_out"),
- SUNXI_FUNCTION(0x2,"nand0")),/* NDQ7 */
+ SUNXI_FUNCTION(0x2,"nand0"),/* NDQ7 */
+ SUNXI_FUNCTION(0x3,"mmc2")),/* D7 */
SUNXI_PIN(SUNXI_PINCTRL_PIN(C,16),
SUNXI_FUNCTION(0x0,"gpio_in"),
SUNXI_FUNCTION(0x1,"gpio_out"),
|
and it recognizes eMMC without problems.
you really don't need these, they dont' work anyway, especially without
modifying the mmc2_pins_a set.
Using dd for performance measure it gives:
|
root@egpr:/mnt# dd if=/dev/zero of=1GBfilebs=1Mcount=1K
1024+0records in
1024+0records out
1073741824bytes (1.1GB)copied,78.7895s,13.6MB/s
root@egpr:/mnt# dd of=/dev/nullif=1GBfile
2097152+0records in
2097152+0records out
1073741824bytes (1.1GB)copied,28.6025s,37.5MB/s
|
Compared to my last results write is the same (13.6 vs. 13.5 MB/s) as
speed but the read is 80% faster (37.5 vs. 20.8 MB/s).
Yeah i got 40 mb/s but i used a blocksize of 16M. So slightly better, as
for writes, i got 17mb/s also slightly better.
As for why it worked now and not before, seems like something went wrong
with the patching of things?
Best regards
Chris
--
You received this message because you are subscribed to the Google Groups
"linux-sunxi" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
