This Gist attempts to document the research and (hopefully) steps to successfully allowing U-Boot to perform an OTA (over-the-air) update.
Please comment with corrections and suggestions!
Performing an OTA Update: Place the contents of the toFlash directory onto the update partition of the Intel Edison. Then, at the U-Boot prompt run do_ota. Or, if the system is running in multi-user mode, run as root: reboot ota.
Motivation: Currently (Yocto 2.1, Intel Edison) has an update partition of ~800MB. The Yocto toFlash directory is ~500MB, so placing it on to the update partition is no problem, allowing an OTA update through U-Boot. However, to do an OTA update using Emutex Labs' Ubilinxu is currently impossible as the Ubilinux toFlash directory is ~1.6GB.
Goal: The Intel Edison allows for an external SD card; the goal is to point the U-Boot do_load_ota_scr env variable to the external SD card instead of the eMMC update partition (mmc 0:9). This is a superior solution to resizing partitions and allowing a larger update partition.
The Intel Edison uses a SoC that is a 22nm Intel Atom "Tangier" (Z34XX) that includes two Atom Silvermont cores (see Wikipedia and Intel's Z34XX Brief). It's based on the "Merrifield" and "Moorefield" families.
The Edison uses a Langwell GPIO interface:
# cat /proc/iomem
ff008000-ff008fff : 0000:00:0c.0
ff008000-ff008fff : langwell_gpioSee section 3.3 Managed NAND (eMMC) Flash of the Intel Edison Compute Module Hardware Guide (Rev 4)
The Intel Edison has a 4GB eMMC device for its filesystem and the U-Boot environment partitions. This can be seen in the output of the parted command below.
# parted /dev/mmcblk0 print
Model: MMC H4G1d (sd/mmc)
Disk /dev/mmcblk0: 3909MB
Sector size (logical/physical): 512B/512B
Partition Table: gpt
Disk Flags:
Number Start End Size File system Name Flags
1 1049kB 3146kB 2097kB u-boot0
2 3146kB 4194kB 1049kB u-boot-env0
3 4194kB 6291kB 2097kB u-boot1
4 6291kB 7340kB 1049kB u-boot-env1
5 7340kB 8389kB 1049kB ext2 factory
6 8389kB 33.6MB 25.2MB panic
7 33.6MB 67.1MB 33.6MB fat16 boot
8 67.1MB 1678MB 1611MB ext4 rootfs
9 1678MB 2483MB 805MB fat32 update
10 2483MB 3909MB 1426MB ext4 home# cat /sys/kernel/debuc/mmc0/ios
clock: 200000000 Hz
actual clock: 200000000 Hz
vdd: 7 (1.65 - 1.95 V)
bus mode: 2 (push-pull)
chip select: 0 (don't care)
power mode: 2 (on)
bus width: 3 (8 bits)
timing spec: 8 (mmc high-speed SDR200)
signal voltage: 0 (1.80 V)See section 4.3 SD Card Interface of the Intel Edison Compute Module Hardware Guide (Rev 4)
There is one external SD card interface, according to the Edison's Hardware Guide. Although, cat /proc/iomem and lspci -vvv reveal three SD Host Controllers:
# cat /proc/iomem
ff3fa000-ff3fa0ff : 0000:00:01.2
ff3fa000-ff3fa0ff : mmc1
ff3fb000-ff3fb0ff : 0000:00:01.3
ff3fb000-ff3fb0ff : mmc2
ff3fc000-ff3fc0ff : 0000:00:01.0
ff3fc000-ff3fc0ff : mmc0# lspci -vvv
00:01.0 SD Host controller: Intel Corporation Device 1190 (rev 01) (prog-if 01)
Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Latency: 64
Interrupt: pin A routed to IRQ 0
Region 0: Memory at ff3fc000 (32-bit, non-prefetchable) [size=256]
Capabilities: [b0] Power Management version 3
Flags: PMEClk- DSI+ D1- D2- AuxCurrent=0mA PME(D0+,D1-,D2-,D3hot+,D3cold-)
Status: D3 NoSoftRst- PME-Enable- DSel=0 DScale=0 PME-
Capabilities: [b8] Vendor Specific Information: Len=08 <?>
Capabilities: [c0] PCI-X non-bridge device
Command: DPERE- ERO+ RBC=512 OST=1
Status: Dev=ff:1f.7 64bit- 133MHz+ SCD- USC- DC=simple DMMRBC=512 DMOST=1 DMCRS=8 RSCEM- 266MHz+ 533MHz-
Capabilities: [100 v1] Vendor Specific Information: ID=0000 Rev=0 Len=024 <?>
Kernel driver in use: sdhci-pci
00:01.2 SD Host controller: Intel Corporation Device 1190 (rev 01) (prog-if 01)
Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Latency: 64
Interrupt: pin B routed to IRQ 37
Region 0: Memory at ff3fa000 (32-bit, non-prefetchable) [size=256]
Capabilities: [b0] Power Management version 3
Flags: PMEClk- DSI+ D1- D2- AuxCurrent=0mA PME(D0+,D1-,D2-,D3hot+,D3cold-)
Status: D3 NoSoftRst- PME-Enable- DSel=0 DScale=0 PME-
Capabilities: [b8] Vendor Specific Information: Len=08 <?>
Capabilities: [c0] PCI-X non-bridge device
Command: DPERE- ERO+ RBC=512 OST=1
Status: Dev=ff:1f.7 64bit- 133MHz+ SCD- USC- DC=simple DMMRBC=512 DMOST=1 DMCRS=8 RSCEM- 266MHz+ 533MHz-
Capabilities: [100 v1] Vendor Specific Information: ID=0000 Rev=0 Len=024 <?>
Kernel driver in use: sdhci-pci
00:01.3 SD Host controller: Intel Corporation Device 1190 (rev 01) (prog-if 01)
Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Latency: 64
Interrupt: pin C routed to IRQ 38
Region 0: Memory at ff3fb000 (32-bit, non-prefetchable) [size=256]
Capabilities: [b0] Power Management version 3
Flags: PMEClk- DSI+ D1- D2- AuxCurrent=0mA PME(D0+,D1-,D2-,D3hot+,D3cold-)
Status: D3 NoSoftRst- PME-Enable- DSel=0 DScale=0 PME-
Capabilities: [b8] Vendor Specific Information: Len=08 <?>
Capabilities: [c0] PCI-X non-bridge device
Command: DPERE- ERO+ RBC=512 OST=1
Status: Dev=ff:1f.7 64bit- 133MHz+ SCD- USC- DC=simple DMMRBC=512 DMOST=1 DMCRS=8 RSCEM- 266MHz+ 533MHz-
Capabilities: [100 v1] Vendor Specific Information: ID=0000 Rev=0 Len=024 <?>
Kernel driver in use: sdhci-pciThe eMMC has been verified to be the mmc0 device at PCI address 0000:00:01.0 and memory address 0xff3fc00. I believe that mmc1 (0xff3fa00) is the external SD card and that mmc2 is a hardware disabled controller (although, I'm still confused about this, as seen in the IRQ section below). The following output from dmesg cross-referenced with lspci -vvv (above) and the device tree in /sys/bus/mmc/drivers/mmcblk seems to verify:
# dmesg | grep mmc
[ 1.595165] mmc1: no vqmmc regulator found
[ 1.595544] mmc1: SDHCI controller on PCI [0000:00:01.2] using ADMA
[ 1.606109] mmc2: no vqmmc regulator found
[ 1.606576] mmc2: SDHCI controller on PCI [0000:00:01.3] using ADMA
[ 1.739385] mmc1: new high speed SDHC card at address aaaa
[ 1.740215] mmcblk1: mmc1:aaaa SL32G 28.7 GiB
[ 1.751294] mmcblk1: p1root@edison:/sys/bus/mmc/drivers/mmcblk# ll
--w------- 1 root root 4.0K Jun 8 15:40 bind
lrwxrwxrwx 1 root root 0 Jun 8 15:40 mmc0:0001 -> ../../../../devices/pci0000:00/0000:00:01.0/mmc_host/mmc0/mmc0:0001
lrwxrwxrwx 1 root root 0 Jun 8 15:40 mmc1:aaaa -> ../../../../devices/pci0000:00/0000:00:01.2/mmc_host/mmc1/mmc1:aaaa
--w------- 1 root root 4.0K Jan 1 2000 uevent
--w------- 1 root root 4.0K Jun 8 15:40 unbindIt can be shown that device mmc1 is at address aaaa which points to pci0000:00/0000:00:01.2 which is at address ff3fa000. This address is important when telling U-Boot where to look for the device.
# cat /sys/kernel/debug/mmc1/ios
clock: 50000000 Hz
actual clock: 50000000 Hz
vdd: 17 (2.9 ~ 3.0 V)
bus mode: 2 (push-pull)
chip select: 0 (don't care)
power mode: 2 (on)
bus width: 2 (4 bits)
timing spec: 2 (sd high-speed)
signal voltage: 0 (3.30 V)In the output below, we have, by colum:
- IRQ (Interrupt Request) number
- Interrupts fired on CPU0
- Interrupts fired on CPU1
- Type of interrupt
- Name of the interrupt
Interrupts are registered in Kernel code using the request_irq function:
int request_irq ( unsigned int irq,
irq_handler_t handler,
unsigned long irqflags,
const char * devname,
void * dev_id);# cat /proc/interrupts | grep 'mmc\|sd_cd\|CPU0'
CPU0 CPU1
0: 5166 0 IO-APIC-fasteoi mmc0
37: 2341 0 IO-APIC-fasteoi mmc1
38: 5612 0 IO-APIC-fasteoi mmc2
320: 175 0 LNW-GPIO-demux bcmsdh_sdmmc
333: 2 0 LNW-GPIO-demux sd_cdThe sd_cd interrupt is coming from the Langwell GPIO block and is for SD Card Detection. Every insert/removal of the SD card causes an interrupt on this handler. The confusing thing is that all three mmc devices have registered interrupts, and in this specific case, mmc2 (which is though to be disabled) has had more interrupts than mmc1. However, after inserting/removing an SD card, mmc1 not mmc2 increments. I am unsure what bcmsdh_sdmmc is, although the bcm portion of the name scares me that it's a proprietary Broadcomm thing.
I have a hunch that mmc2 might be a SDIO device -- possibly the wlan?
# dmesg | grep -i gpio
[ 1.595518] sdhci-pci 0000:00:01.2: GPIO: 77
[ 1.595541] sdhci-pci 0000:00:01.2: request_irq(333, handler, 4d00000003, sd_cd_parker, slot);Digging through the U-Boot source code after the edison-src/meta-intel-edison/meta-intel-edison-bsp/recipes-bsp/u-boot/files/upstream_to_edison.path has been applied shows promising results. It seems like support for the external mmc1 can be enabled by adding some extra code.
Starting from boards.cfg we will see how the eMMC (device mmc0) is being enabled:
# edison-src/out/current/build/tmp/work/edison-poky-linux/u-boot/2014.04-1-r0/git/boards.cfg
#Status, Arch, CPU:SPLCPU, SoC, Vendor, Board name, Target, Options, Maintainers
Active x86 sivermont tangier intel edison edison edison:SYS_USB_OTG_BASE=0xf9100000,SYS_EMMC_PORT_BASE=0xff3fc000,SYS_TEXT_BASE=0x1101000 -This config file shows that CONFIG_SYS_EMMC_PORT_BASE=0xff3fc000, which is used later in tangier.c:board_mmc_init.
// edison-src/out/current/build/tmp/work/edison-poky-linux/u-boot/2014.04-1-r0/git/include/configs/edison.h
/*
* MMC
*/
#define CONFIG_MD5
#define CONFIG_GENERIC_MMC
#define CONFIG_MMC
#define CONFIG_SDHCI
#define CONFIG_TANGIER_SDHCI
#define CONFIG_CMD_MMC
#define CONFIG_MMC_SDMA
/*#define CONFIG_MMC_TRACE*/This header config file seems to include support for MMC and the TANGIER_SDHCI. Uncommenting CONFIG_MMC_TRACE will print out the SD/MMC commands and show the contents of the registers from the MMC/SD device.
// edison-src/out/current/build/tmp/work/edison-poky-linux/u-boot/2014.04-1-r0/git/arch/x86/cpu/tangier/tangier.c
int board_mmc_init(bd_t * bis)
{
int index = 0;
unsigned int base = CONFIG_SYS_EMMC_PORT_BASE + (0x40000 * index);
return tangier_sdhci_init(base, index, 4);
}This code initializes the eMMC. It is odd that 0x40000 is being multiplied by the index (which is 0 in this case). One would think duplicating this code with index=1 would add the next device, but it doesn't seem to work.
I changed the tangier.c:board_mmc_init function as follows:
int board_mmc_init(bd_t * bis)
{
int index = 0;
unsigned int base = CONFIG_SYS_EMMC_PORT_BASE + (0x40000 * index);
// eMMC at 0cff3fc000
tangier_sdhci_init(base, index, 4)
// mmc1 external SD card, index=1, width=4
parker_sdhci_init(0xff3fa000, 1, 4);
return 0;
}Then, I copied and pasted the tangier_sdhci_init function and renamed to parker_sdhci_init in drivers/mmc/tangier_sdhci.c:
int parker_sdhci_init(u32 regbase, int index, int bus_width)
{
struct sdhci_host *host = NULL;
host = (struct sdhci_host *)malloc(sizeof(struct sdhci_host));
if (!host) {
printf("sdhci__host malloc fail!\n");
return 1;
}
memset(host, 0x00, sizeof(struct sdhci_host));
host->name = "parker_sdhci";
host->ioaddr = (void *)regbase;
host->quirks = SDHCI_QUIRK_NO_HISPD_BIT | SDHCI_QUIRK_BROKEN_VOLTAGE |
SDHCI_QUIRK_BROKEN_R1B | SDHCI_QUIRK_32BIT_DMA_ADDR |
SDHCI_QUIRK_WAIT_SEND_CMD;
host->voltages = MMC_VDD_29_30; // | MMC_VDD_32_33 | MMC_VDD_33_34 | MMC_VDD_165_195;
printf("About to run sdhci_readw.\n");
// All this function does is dereferrences the value at 0xff3fa0fe (which is the SDHC version reg)
host->version = sdhci_readw(host, SDHCI_HOST_VERSION);
// I am getting back 0xffff for version!
// I should be expecting the lower 8 bits to be 0x02 (SD v3.0)
printf("Version returned: %d\n", host->version);
host->index = index;
host->host_caps = MMC_MODE_HC;
// Added to see if bus_width needed to be 4 bits (as per docs)
if (bus_width == 4)
host->host_caps |= MMC_MODE_4BIT;
printf("About to run add_sdhci...\n");
// Changed MAX clock to 50MHz
add_sdhci(host, 50000000, 400000);
printf("Ran add_sdhci.\n");
return 0;
}When I upload and run, U-Boot will hang and then restart.
Have you seen this one: https://github.com/01org/edison-u-boot/commits/edison-v2015.10 ?I had forked one of the branches from there and started cleaning up and upstreaming. We are now just 8 patches far from out-of-the-box support for Edison in U-Boot.