system hangs or kernel panic when android system is loading on i.MX6 Solo board

[lqw39]

We're debugging Solo (not sololite) board, and meeting a critical problem: system always hang or kernel panic (Oops) when loading android system.

It seems that this problem is caused by surfaceflinger service. kernel can run normally if surfaceflinger removed in init.rc.

kernel panic info:

[<c0095044>] (wq_worker_waking_up+0xc/0x58) from [<c007a480>] (ttwu_do_activate.clone.0+0x2c/0x40)

[<c007a480>] (ttwu_do_activate.clone.0+0x2c/0x40) from [<c007a768>] (try_to_wake_up+0x2d4/0x324)

[<c007a768>] (try_to_wake_up+0x2d4/0x324) from [<c009302c>] (__queue_work+0x150/0x384)

[<c009302c>] (__queue_work+0x150/0x384) from [<c0089b08>] (run_timer_softirq+0x124/0x244)

[<c0089b08>] (run_timer_softirq+0x124/0x244) from [<c0083c18>] (__do_softirq+0xa8/0x170)

[<c0083c18>] (__do_softirq+0xa8/0x170) from [<c0083f2c>] (irq_exit+0x90/0x98)

[<c0083f2c>] (irq_exit+0x90/0x98) from [<c0049fe4>] (handle_IRQ+0x50/0xac)

[<c0049fe4>] (handle_IRQ+0x50/0xac) from [<c004900c>] (__irq_svc+0x4c/0xe8)

[<c004900c>] (__irq_svc+0x4c/0xe8) from [<c05ac500>] (_raw_spin_unlock_irq+0x18/0x40)

[<c05ac500>] (_raw_spin_unlock_irq+0x18/0x40) from [<c004d194>] (die+0xbc/0x2a4)

[<c004d194>] (die+0xbc/0x2a4) from [<c0053e98>] (__do_kernel_fault+0x64/0x84)

[<c0053e98>] (__do_kernel_fault+0x64/0x84) from [<c0054038>] (do_page_fault+0x180/0x2e0)

[<c0054038>] (do_page_fault+0x180/0x2e0) from [<c0043464>] (do_DataAbort+0x34/0x98)

[<c0043464>] (do_DataAbort+0x34/0x98) from [<c0048f90>] (__dabt_svc+0x70/0xa0)

Exception stack(0xd5157ae0 to 0xd5157b28)

Or,

[<c0040d50>] (__dabt_svc+0x70/0xa0) from [<c00617e4>] (mxc_timer_interrupt+0x20/0x38)

[<c00617e4>] (mxc_timer_interrupt+0x20/0x38) from [<c00ab9fc>] (handle_irq_event_percpu+0x50/0x19c)

[<c00ab9fc>] (handle_irq_event_percpu+0x50/0x19c) from [<c00abb9c>] (handle_irq_event+0x54/0x84)

[<c00abb9c>] (handle_irq_event+0x54/0x84) from [<c00adc9c>] (handle_fasteoi_irq+0x84/0x138)

[<c00adc9c>] (handle_fasteoi_irq+0x84/0x138) from [<c00ab634>] (generic_handle_irq+0x30/0x3c)

[<c00ab634>] (generic_handle_irq+0x30/0x3c) from [<c0041b7c>] (handle_IRQ+0x30/0x84)

[<c0041b7c>] (handle_IRQ+0x30/0x84) from [<c0040dcc>] (__irq_svc+0x4c/0xa8)

[<c0040dcc>] (__irq_svc+0x4c/0xa8) from [<c0044c28>] (die+0xbc/0x2c0)

[<c0044c28>] (die+0xbc/0x2c0) from [<c004a8e8>] (__do_kernel_fault+0x64/0x84)

[<c004a8e8>] (__do_kernel_fault+0x64/0x84) from [<c004aa88>] (do_page_fault+0x180/0x2e0)

[<c004aa88>] (do_page_fault+0x180/0x2e0) from [<c003b240>] (do_DataAbort+0x34/0x98)

[<c003b240>] (do_DataAbort+0x34/0x98) from [<c0040d50>] (__dabt_svc+0x70/0xa0)

Have you ever encountered this?

[jimmychan]

What version of the BSP are you using?

[lqw39]

We are using the 13.4.1 anroid release, and I'm fairly certain it's due to DDR settings.

DDR connecting lead on our board is different from referenc design:

Our board:

Chip1 – CS0/D0-D15

Chip2 – CS0/D16-D31

Chip3 – CS1/D0-D15

Chip4 – CS1/D16-D31

Reference board:

Chip1 – CS0/D0-D15

Chip2 – CS0/D16-D31

Chip3 – CS0/D32-D47

Chip4 – CS0/D48-D63

So we need enable CS1 in flash_header.S, but I'm not clear how to config MMDC1.

[JackyAtFreescal · Accepted solution]

DSIZ bits in MDCTL register is needed to change.

Extracted from IMX6SL RM:

The following registers in the MMDC define the DDR address space:

• MDMISC[DDR_4_BANK]—Defines either 4 or 8 banks in the DDR device

• MDCTL[DSIZ]—Defines the DDR data bus width of x16, x32 or x64

• MDMISC[BI]—Defines whether bank interleaving is on or off

• MDCTL[COL]—Defines the column size of the DDR device

• MDCTL[ROW]—Defines the row size of the DDR device

[jimmychan]

I will try to ask someone in expert team to help you.