[PATCH V2] cmd: mem: fix to display wrong memory information
When run meminfo command, it's displayed wrong memory information. Because some boards are that gd->ram_size is reassigned to other value in board file. Additionally, display a memory bank information.
On 4G RPI4 target - Before U-Boot> meminfo DRAM: 948MiB - After U-Boot> meminfo Bank #0: 0 948 MiB Bank #1: 40000000 2.9 GiB Bank #2: 0 0 Bytes Bank #3: 0 0 Bytes DRAM: 3.9GiB
Signed-off-by: Jaehoon Chung jh80.chung@samsung.com --- Changes in v2: - Change patch subject prefix from "common: board_f" to "cmd: mem" - Update commit-msg - Revert common/board_f.c modification. Instead, add codes in mem.c --- cmd/mem.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-)
diff --git a/cmd/mem.c b/cmd/mem.c index 1d4f2bab2f9a..86f48a6e121a 100644 --- a/cmd/mem.c +++ b/cmd/mem.c @@ -1387,8 +1387,22 @@ U_BOOT_CMD( static int do_mem_info(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[]) { + unsigned long long size; + +#ifdef CONFIG_NR_DRAM_BANKS + int i; + + for (i = size = 0; i < CONFIG_NR_DRAM_BANKS; i++) { + size += gd->bd->bi_dram[i].size; + printf("Bank #%d: %llx ", i, + (unsigned long long)(gd->bd->bi_dram[i].start)); + print_size(gd->bd->bi_dram[i].size, "\n"); + } +#else + size = gd->ram_size; +#endif puts("DRAM: "); - print_size(gd->ram_size, "\n"); + print_size(size, "\n");
return 0; }
On 28.01.21 03:35, Jaehoon Chung wrote:
When run meminfo command, it's displayed wrong memory information. Because some boards are that gd->ram_size is reassigned to other value in board file. Additionally, display a memory bank information.
On 4G RPI4 target
- Before U-Boot> meminfo DRAM: 948MiB
- After U-Boot> meminfo Bank #0: 0 948 MiB Bank #1: 40000000 2.9 GiB Bank #2: 0 0 Bytes Bank #3: 0 0 Bytes DRAM: 3.9GiB
Signed-off-by: Jaehoon Chung jh80.chung@samsung.com
Changes in v2:
- Change patch subject prefix from "common: board_f" to "cmd: mem"
- Update commit-msg
- Revert common/board_f.c modification. Instead, add codes in mem.c
cmd/mem.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-)
diff --git a/cmd/mem.c b/cmd/mem.c index 1d4f2bab2f9a..86f48a6e121a 100644 --- a/cmd/mem.c +++ b/cmd/mem.c @@ -1387,8 +1387,22 @@ U_BOOT_CMD( static int do_mem_info(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[]) {
- unsigned long long size;
+#ifdef CONFIG_NR_DRAM_BANKS
- int i;
- for (i = size = 0; i < CONFIG_NR_DRAM_BANKS; i++) {
size += gd->bd->bi_dram[i].size;printf("Bank #%d: %llx ", i,(unsigned long long)(gd->bd->bi_dram[i].start));print_size(gd->bd->bi_dram[i].size, "\n");- }
+#else
- size = gd->ram_size;
+#endif
CONFIG_NR_DRAM_BANKS is always defined! There has been some work to all #ifdef's in this area, as it's not needed. So please don't add new #ifdef's here, as they are not needed. Please search for:
Remove CONFIG_NR_DRAM_BANKS option and bi_memstart/memsize from bd_info
If your board does not display the correct size, then you most likely have some issues in your local dram_init_banksize() or dram_init() implementation. If the total RAM cannot be used by U-Boot (which is a common issue) then please use something like board_get_usable_ram_top() and/or get_effective_memsize() for this.
Here on my 4GiB Octeon MIPS board:
... DRAM: 256 MiB (4 GiB total) ... => meminfo DRAM: 4 GiB => bdinfo boot_params = 0x0000000000000000 DRAM bank = 0x0000000000000000 -> start = 0xffffffff80000000 -> size = 0x0000000010000000 ...
HTH.
Thanks, Stefan
On 1/28/21 6:33 PM, Stefan Roese wrote:
On 28.01.21 03:35, Jaehoon Chung wrote:
When run meminfo command, it's displayed wrong memory information. Because some boards are that gd->ram_size is reassigned to other value in board file. Additionally, display a memory bank information.
On 4G RPI4 target
- Before
U-Boot> meminfo DRAM: 948MiB
- After
U-Boot> meminfo Bank #0: 0 948 MiB Bank #1: 40000000 2.9 GiB Bank #2: 0 0 Bytes Bank #3: 0 0 Bytes DRAM: 3.9GiB
Signed-off-by: Jaehoon Chung jh80.chung@samsung.com
Changes in v2:
- Change patch subject prefix from "common: board_f" to "cmd: mem"
- Update commit-msg
- Revert common/board_f.c modification. Instead, add codes in mem.c
cmd/mem.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-)
diff --git a/cmd/mem.c b/cmd/mem.c index 1d4f2bab2f9a..86f48a6e121a 100644 --- a/cmd/mem.c +++ b/cmd/mem.c @@ -1387,8 +1387,22 @@ U_BOOT_CMD( static int do_mem_info(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[]) { + unsigned long long size;
+#ifdef CONFIG_NR_DRAM_BANKS + int i;
+ for (i = size = 0; i < CONFIG_NR_DRAM_BANKS; i++) { + size += gd->bd->bi_dram[i].size; + printf("Bank #%d: %llx ", i, + (unsigned long long)(gd->bd->bi_dram[i].start)); + print_size(gd->bd->bi_dram[i].size, "\n"); + } +#else + size = gd->ram_size; +#endif
CONFIG_NR_DRAM_BANKS is always defined! There has been some work to all #ifdef's in this area, as it's not needed. So please don't add new #ifdef's here, as they are not needed. Please search for:
Remove CONFIG_NR_DRAM_BANKS option and bi_memstart/memsize from bd_info
If your board does not display the correct size, then you most likely have some issues in your local dram_init_banksize() or dram_init() implementation. If the total RAM cannot be used by U-Boot (which is a common issue) then please use something like board_get_usable_ram_top() and/or get_effective_memsize() for this.
I understood what you said. Frankly, i had tried to use memory information into my bootscript. So i have checked mmcinfo command.
Right, there is included information not to use by U-boot side.
I thought a issue what not display Total Ram information with meminfo. (Because there isn't any description about meminfo.) But if it's displayed memory that can be used by U-boot, this patch doesn't need.
I have tested on my own targets. (RPI4 2G/4G/8G, meson target - VIM3(4G), Odroid-N2(4G)) - Those boards were displayed almost under 1G.
Thanks for comment and checking!
Best Regards, Jaehoon Chung
Here on my 4GiB Octeon MIPS board:
... DRAM: 256 MiB (4 GiB total) ... => meminfo DRAM: 4 GiB => bdinfo boot_params = 0x0000000000000000 DRAM bank = 0x0000000000000000 -> start = 0xffffffff80000000 -> size = 0x0000000010000000 ...
HTH.
Thanks, Stefan
participants (2)
-
Jaehoon Chung -
Stefan Roese