DRAM Controller/Quirks

From linux-sunxi.org
Jump to navigation Jump to search

This is a list of some strange things and unresolved mysteries.

Suspicious GPS manipulations

The current code in upstream u-boot: http://git.denx.de/?p=u-boot.git;a=blob;f=arch/arm/cpu/armv7/sunxi/dram.c;h=b43c4b41d3c53dc339b3f4f4229daa9fca7f9764;hb=3fe1a8545b55d31a6db2d9e60d962c4f6e048913#l218

The first commit introducing this code in the u-boot-sunxi says "if we don't reset the gps module, it will access sdram but sdram is not ready, and the system will die...".

The corresponding reference in the Allwinner boot0 bootloader is yet to be found.

SDR_SCSR register ('csel' in the sunxi_dram_reg struct) on Allwinner A10

The current code in upstream u-boot: not available yet.

The code in Allwinner boot1 bootloader: https://github.com/hno/allwinner-boot/blob/6fd439377f0f0f0305d61fe6b87c9e77666facb3/boot1/core/standby/dram_standby.c#L326

Unless a magic value 0x16237495 is written to this register on Allwinner A10, the DRAM controller fails to operate properly. This register always reads back as 0.


The current code in upstream u-boot: http://git.denx.de/?p=u-boot.git;a=blob;f=arch/arm/cpu/armv7/sunxi/dram.c;h=b43c4b41d3c53dc339b3f4f4229daa9fca7f9764;hb=3fe1a8545b55d31a6db2d9e60d962c4f6e048913#l535

The same code in boot0: https://github.com/hno/allwinner-boot/blob/70576dd51b65722abadd5ea6169f6c20639cf211/boot0/drv/init_dram/dram_init.c#L409

We can see, that this particular code got mangled on the way to the upstream u-boot and does not make any sense there. For example:

  • in the original boot0 code we had "mctl_write_w(SDR_DPCR, 0x16510000)" (write to the register) and in upstream u-boot it now looks like "setbits_le32(&dram->ppwrsctl, 0x16510000)" (set bits in the register)
  • in the original boot0 code it was issuing three commands "0x12, 0x17, 0x13" (Self-Refresh entry, Self-Refresh exit, Refresh), but in the upstream u-boot they have become "0x12, 0x12, 0x13" (Self-Refresh entry, Self-Refresh entry, Refresh)

It is very likely that the original code had something to do with wake up from super-standby (which keeps DDR3 in Self-Refresh mode?), where the previously stored ZQ calibration settings are retrieved from the battery backed Timer General Purpose Register 0 in the RTC power domain.

Some messy commits in the Allwinner bootloader: