Ps3 Memory leaks... (Developing OpenBOR port for PS3)

I did a sample, to test if it crashs if it malloc a too big size of data :

Code: 
typedef struct {
    uint32_t total;
    uint32_t avail;
} sysMemoryInfo_t;

s32 sysMemoryGetUserMemorySize(sysMemoryInfo_t * MemInfo)
{
    lv2syscall1(SYSCALL_MEMORY_GET_USER_MEMORY_SIZE, (uint64_t) MemInfo);
    return_to_user_prog(s32);
}

int main()
{
    FILE *fp;
    char str[128];
    u64 size = 0;
    sysMemoryInfo_t MemInfo;
    u8 *test;
   
    fp = fopen("/dev_hdd0/tmp/log.txt", "w");
    if(fp==NULL) return 0;
   
    while(1)
    {
        size += 0x100000;
        memset(&MemInfo, 0, sizeof(sysMemoryInfo_t));
        sysMemoryGetUserMemorySize(&MemInfo);
        sprintf(str, "MEM : %X / %X\n", (unsigned int) MemInfo.avail, (unsigned int) MemInfo.total); fputs(str, fp);
        sprintf(str, "Malloc : %lX", (unsigned long int) size); fputs(str, fp); usleep(100);
        test = (u8 *) malloc(size);
        if(test==NULL) {
            strcpy(str, " FAILED \n"); fputs(str, fp);
            sysMemoryGetUserMemorySize(&MemInfo);
            sprintf(str, "NEW MEM : %X / %X\n", (unsigned int) MemInfo.avail, (unsigned int) MemInfo.total); fputs(str, fp); usleep(100);
            fclose(fp);
            return 0;
        } else {
            strcpy(str, " OK \n"); fputs(str, fp);
            sysMemoryGetUserMemorySize(&MemInfo);
            sprintf(str, "NEW MEM : %X / %X\n", (unsigned int) MemInfo.avail, (unsigned int) MemInfo.total); fputs(str, fp); usleep(100);
        }
       
        free(test);
    }
}

and it works fine... here the last loops logs :

Code: 
MEM : EAF0000 / EC00000
Malloc : EA00000 OK
NEW MEM : F0000 / EC00000
MEM : EAF0000 / EC00000
Malloc : EB00000 FAILED
NEW MEM : EAF0000 / EC00000

I don't get it ><
 
Zar said:
I did a sample, to test if it crashs if it malloc a too big size of data :

Code: 
typedef struct {
    uint32_t total;
    uint32_t avail;
} sysMemoryInfo_t;

s32 sysMemoryGetUserMemorySize(sysMemoryInfo_t * MemInfo)
{
    lv2syscall1(SYSCALL_MEMORY_GET_USER_MEMORY_SIZE, (uint64_t) MemInfo);
    return_to_user_prog(s32);
}

int main()
{
    FILE *fp;
    char str[128];
    u64 size = 0;
    sysMemoryInfo_t MemInfo;
    u8 *test;
 
    fp = fopen("/dev_hdd0/tmp/log.txt", "w");
    if(fp==NULL) return 0;
 
    while(1)
    {
        size += 0x100000;
        memset(&MemInfo, 0, sizeof(sysMemoryInfo_t));
        sysMemoryGetUserMemorySize(&MemInfo);
        sprintf(str, "MEM : %X / %X\n", (unsigned int) MemInfo.avail, (unsigned int) MemInfo.total); fputs(str, fp);
        sprintf(str, "Malloc : %lX", (unsigned long int) size); fputs(str, fp); usleep(100);
        test = (u8 *) malloc(size);
        if(test==NULL) {
            strcpy(str, " FAILED \n"); fputs(str, fp);
            sysMemoryGetUserMemorySize(&MemInfo);
            sprintf(str, "NEW MEM : %X / %X\n", (unsigned int) MemInfo.avail, (unsigned int) MemInfo.total); fputs(str, fp); usleep(100);
            fclose(fp);
            return 0;
        } else {
            strcpy(str, " OK \n"); fputs(str, fp);
            sysMemoryGetUserMemorySize(&MemInfo);
            sprintf(str, "NEW MEM : %X / %X\n", (unsigned int) MemInfo.avail, (unsigned int) MemInfo.total); fputs(str, fp); usleep(100);
        }
     
        free(test);
    }
}

and it works fine... here the last loops logs :

Code: 
MEM : EAF0000 / EC00000
Malloc : EA00000 OK
NEW MEM : F0000 / EC00000
MEM : EAF0000 / EC00000
Malloc : EB00000 FAILED
NEW MEM : EAF0000 / EC00000

I don't get it ><
Click to expand...

Mmm... Big thanks for your test. I changed the script library to use a char * instead of fixed length arrays.
Well ram usage is more large : about 180mb against ~130mb. Used a linked list to memorize into stack the potential malloc memory.
I don't know what is the issue...
The ps3 emulator works well until 200mb of ram... Now I think the recursion is the problem. Mmm..
 
Last edited:
Rewritte my dmalloc library and 0 memory leak found!!
So it works well..
Just I need if RAM is full for a last attempt...
 
Hi to all, I'm studying all parser/lexer code.
I found that used ram is just 28MB when the ps3 freezes.
In this way I can't debug.
How can I debug it?
The strange thing is taht on ps3 emulator it works well, but on a real ps3 it freezes.
Please help...
I suspect an illegal memory access.
into preprocessorlib folder and into pp_lexer.c there is the function pp_lexer_GetNextToken().
The app crashes before re enter in this func.
Can you help me?
 
I can confirm that into preprocessorlib/pp_parser.c
there is the function HRESULT pp_parser_lex_token(pp_parser *self, bool skip_whitespace)
after some loops the instruction step is freezed before
success = (int)SUCCEEDED(pp_lexer_GetNextToken(&self->lexer, &self->token));
and never enters into pp_lexer_GetNextToken()

I searched for a memory violation into self->lexer, self->token, self addresses but it seems all ok...
why?
Is it right if I think that there is some recursion limit or... some root limits?
 
Zar said:
The link of kozarovv talk about the stacksize, it made me think about SYS_PROCESS_PARAM it allloow you to define the stack sizes maybe if you change these paramaters it will allow you to fix it. hmm, did you define it ?

It's just an idea... https://github.com/Estwald/PSDK3v2/blob/master/psl1ght/ppu/include/sys/process.h
Click to expand...

Thanks for your answer.
I defined my custom stack resize directive

#define SYS_PROCESS_PARAM_OPENBOR(prio,stacksize) \
sys_process_param_t __sys_process_param __attribute__((aligned(8), section(".sys_proc_param"), unused)) = { \
sizeof(sys_process_param_t), \
SYS_PROCESS_SPAWN_MAGIC, \
SYS_PROCESS_SPAWN_VERSION_330, \
SYS_PROCESS_SPAWN_FW_VERSION_330, \
prio, \
stacksize, \
SYS_PROCESS_SPAWN_MALLOC_PAGE_SIZE_1M, \
SYS_PROCESS_SPAWN_PPC_SEG_DEFAULT\
};
SYS_PROCESS_PARAM_OPENBOR(1001, 0x1F000000)


But no luck..
I don't know if I wrote it well...
kozarovv said:
https://stackoverflow.com/questions/2630054/does-c-limit-recursion-depth
Click to expand...

Thanks! I tested it but I think that for ps3 I need some other func.. That is for Linux.
Maybe the setrlimt? If I can set it on ps3..


https://stackoverflow.com/questions...trlimit-and-provoke-a-stack-overflow-segfault
 
EUREKAAAAA!!
I substiture all infinite loops with a for with limited loop count definited into a macro + SYS_PROCESS_PARAM(1001, SYS_PROCESS_SPAWN_MALLOC_PAGE_SIZE_1M)!!
Not other configs!
Now it works perfectly!!
many nested while(1) into recursive functions are not accepted by PS3 architecture (or PPU??).
Maybe because the processor can't calculate all potential depth...
What do you think about it?
Now I'm ready to release it!
Have you a more elegant method to substitute while(1)?
Thanks!
 
Last edited:
I'm glad you found a solution to ur issue. My suggestion ( 'while(1)' ) was a making it worse, :sem blush: sry.

Have you a more elegant method to substitute while(1)?
Click to expand...
if it's working, just don't change anything ;)

I don't think the spawn_malloc_page value is supposed to be used here. its value is 0x100000
 
White Dragon said:
EUREKAAAAA!!
I substiture all infinite loops with a for with limited loop count definited into a macro + SYS_PROCESS_PARAM(1001, SYS_PROCESS_SPAWN_MALLOC_PAGE_SIZE_1M)!!
Not other configs!
Now it works perfectly!!
many nested while(1) into recursive functions are not accepted by PS3 architecture (or PPU??).
Maybe because the processor can't calculate all potential depth...
What do you think about it?
Now I'm ready to release it!
Have you a more elegant method to substitute while(1)?
Thanks!
Click to expand...

To the staff: please give him then dev status ;)
 
Zar said:
I'm glad you found a solution to ur issue. My suggestion ( 'while(1)' ) was a making it worse, :sem blush: sry.


if it's working, just don't change anything ;)

I don't think the spawn_malloc_page value is supposed to be used here. its value is 0x100000
Click to expand...

0x100000 is 1Mb so it's right.
Better write 0x100000 constant directly maybe.
 
You must log in or register to reply here.

Similar threads

Back
Top