Hello TI-Team.
I have currently one PCIe Root Complex connected to an PCie End Point with a Male-To-Male PCIe cable (both boards are the TMDX654IDKEVM and are running with the Linux SDK).
I am able to write data from RC side into a resource space of the EP. Now i was wondering how can i access on the data i have written on EP side. Is there a common way to read/write data on EP side? I have tried to access the /dev/mem, but i was not able to map the memory.
Thank you for your help!
Felix Windt
Hey Tanmay,
thank you for your answer.
1. I'm using the 07.03 linux-sdk version.
2.1
int PAGE_SIZE = 1;
#define MAP_SIZE 1UL
#define MAP_MASK (MAP_SIZE - 1)
int main(int argc, char *argv[]) {
void *map_base;
off_t target = 0x10200000;
int mem_fd = open("/dev/mem", O_RDWR | O_SYNC);
map_base = mmap(NULL, PAGE_SIZE, PROT_READ | PROT_WRITE, MAP_SHARED, mem_fd, target);
if((unsigned long )map_base == -1) //prevent from Segmentaion fault
{
printf("mmap can't map memory\n");
return 1;
}
printf("Memory mapped at address %p. \n", map_base);
fflush(stdout);
uint16_t* memory = (uint16_t*) map_base;
printf("Value at address 0x%x: 0x%x\n", (unsigned int) target, memory[0]);
fflush(stdout);
munmap (map_base, MAP_SIZE);
return 0;
}
2.2 On some addresses i get an "Unhandled Exception in EL3".
Unhandled Exception in EL3. x30 = 0x00000000700054c0 x0 = 0x0000ffff83af0400 x1 = 0x0000ffff83af0000 x2 = 0x0000000000000000 x3 = 0x0000000000000000 x4 = 0x00000000004009ec x5 = 0x00000000063f1eca x6 = 0x000000000000000a x7 = 0x000000000000000a x8 = 0x0000000000000040 x9 = 0xffffff80ffffffc8 x10 = 0x0000000000000000 x11 = 0x0000ffffdbecc450 x12 = 0x0000000000000001 x13 = 0x0000000000000010 x14 = 0x0000000000000000 x15 = 0x0000000000000000 x16 = 0x0000000000411030 x17 = 0x0000ffff836d8528 x18 = 0x0000000000000000 x19 = 0x00000000004008f0 x20 = 0x0000000000000000 x21 = 0x00000000004006b0 x22 = 0x0000000000000000 x23 = 0x0000000000000000 x24 = 0x0000000000000000 x25 = 0x0000000000000000 x26 = 0x0000000000000000 x27 = 0x0000000000000000 x28 = 0x0000000000000000 x29 = 0x0000ffffdbecc560 scr_el3 = 0x000000000000073d sctlr_el3 = 0x0000000030cd183f cptr_el3 = 0x0000000000000000 tcr_el3 = 0x0000000080803520 daif = 0x00000000000002c0 mair_el3 = 0x00000000004404ff spsr_el3 = 0x0000000080000000 elr_el3 = 0x0000000000400858 ttbr0_el3 = 0x0000000070010b00 esr_el3 = 0x0000000092000010 far_el3 = 0x0000ffff83af0400 spsr_el1 = 0x0000000020000000 elr_el1 = 0x0000ffff8374e3d0 spsr_abt = 0x0000000000000000 spsr_und = 0x0000000000000000 spsr_irq = 0x0000000000000000 spsr_fiq = 0x0000000000000000 sctlr_el1 = 0x0000000034d4d91d actlr_el1 = 0x0000000000000000 cpacr_el1 = 0x0000000000300000 csselr_el1 = 0x0000000000000000 sp_el1 = 0xffff800014110000 esr_el1 = 0x0000000056000000 ttbr0_el1 = 0x00000008c9160000 ttbr1_el1 = 0x0cb6000080e00000 mair_el1 = 0x0000bbff440c0400 amair_el1 = 0x0000000000000000 tcr_el1 = 0x00000032f5507510 tpidr_el1 = 0xffff80086ebf0000 tpidr_el0 = 0x0000ffff83b54400 tpidrro_el0 = 0x0000000000000000 par_el1 = 0x0000000000000000 mpidr_el1 = 0x0000000080000100 afsr0_el1 = 0x0000000000000000 afsr1_el1 = 0x0000000000000000 contextidr_el1 = 0x0000000000000000 vbar_el1 = 0xffff800010081800 cntp_ctl_el0 = 0x0000000000000005 cntp_cval_el0 = 0x0000007f83e09994 cntv_ctl_el0 = 0x0000000000000000 cntv_cval_el0 = 0x0000000000000000 cntkctl_el1 = 0x00000000000000d6 sp_el0 = 0x0000ffffdbecc560 isr_el1 = 0x0000000000000040 dacr32_el2 = 0x0000000000000000 ifsr32_el2 = 0x0000000000000000 cpuectlr_el1 = 0x0000000000000040 cpumerrsr_el1 = 0x000000001a10076e l2merrsr_el1 = 0x000000001320f5f0 cpuactlr_el1 = 0x00001000090ca000
3. "CONFIG_STRICT_DEVMEM" is set to "y"
4. "devmem" does not exist on the board and with "hexdump" i get the following message.
Regards,
Felix
Hi Tanmay,
the devmem2 command exists. Thank you for the info.
We were able to read into the BAR's of the Endpoint in his own physical memory, but the output was always 0xFFFFFFFF. (we also tried the other BAR addresses) Even if we wrote something beforehand from RC side or EP side into one of the BARs(shown in the picture down below), we were able to read nothing more than 0xFFFFFFFF.
Our BAR addresses for the EP (found on RC side):
Regards,
Felix
Felix,
I am trying to follow the discussion above. Could you try the following steps:
1. In RC side, issue the following command on Linux prompt:
/usr/sbin/lspci -vvv
This will list EP's exposed memory map so we can write/read from RC.
2. Further you can use devmem2 utility in RC's linux command prompt, to manually probe EP's BAR settings. Note that devmem2 utility directly addresses physical memory so no conversion from virtual memory is needed. In below example, I am probing EP's BAR register by writing all F's, then read back, that will tell the EP's BAR mask setting. Note that PCIe BAR0 of EP is at offset 0x10 of the PCIe structure.
Note that Step 2 is not needed if the EP is already setup correctly, where we can see the correct writable EP memory region address via lcpsi utility.
regards
Jian
