Hi Wolfgang,
On Wed, Jan 26, 2011 at 5:32 AM, Wolfgang Denk wd@denx.de wrote:
Dear Lei Wen,
In message 1294632087-8025-3-git-send-email-leiwen@marvell.com you wrote:
Pantheon Family processors are highly integrated SoCs based on Sheeva_88SV331x-v5 PJ1 cpu core. Ref: http://www.marvell.com/products/processors/communications/marvell_pantheon_9...
SoC versions Supported:
- PANTHEON920 (TD)
- PANTHEON910 (TTC)
Signed-off-by: Lei Wen leiwen@marvell.com
...
+int dram_init(void) +{
...
- for (; i < CONFIG_NR_DRAM_BANKS; i++) {
- /* If above loop terminated prematurely, we need to set
- * remaining banks' start address & size as 0. Otherwise other
- * u-boot functions and Linux kernel gets wrong values which
- * could result in crash */
Incorrect multiline comment style.
This already fix in the v6 patch... http://patchwork.ozlabs.org/patch/80305/
+/* For preventing risk of instability in reading counter value,
- first set read request to register cvwr and then read same
- register after it captures counter value.
- */
Ditto. Please fix globally.
- writel(COUNT_RD_REQ, &panthtimers->cvwr);
- while (loop--);
Please write:
while (loop--) ;
Fixed...
But then - are you sure the compiler does not optimize this out? You probably want to use __udelay() instead.
From the practice, we think this loop is enough to make timer stablize...
Involve the __udelay() may not suitable for the timer functions...
...
--- /dev/null +++ b/arch/arm/include/asm/arch-pantheon/config.h @@ -0,0 +1,44 @@
...
+/*
- There is no internal RAM in ARMADA100, using DRAM
- TBD: dcache to be used for this
- */
+#define CONFIG_SYS_INIT_SP_ADDR (CONFIG_SYS_TEXT_BASE - 0x00200000) +#define CONFIG_NR_DRAM_BANKS_MAX 2
This looks like board specific code that should not be here.
Yep... I would update the patch for it. For V7...
...
+struct panthmpmu_registers {
- u8 pad0[0x0024];
- u32 ccgr; /*0x0024*/
- u8 pad1[0x0200 - 0x024 - 4];
- u32 wdtpcr; /*0x0200*/
- u8 pad2[0x1020 - 0x200 - 4];
- u32 aprr; /*0x1020*/
- u32 acgr; /*0x1024*/
+};
Please use TAB for vertical alignment of variable names. Please fix globally.
In V6 patch , I think I already do like using tab. :)
Best regards, Lei