Re: [PATCH v1 1/1] memory: export symbols for process memory related functions

[Wei-chin Tsai (蔡維晉)]

On Fri, 2023-06-09 at 12:20 +0100, Russell King (Oracle) wrote:
>  	 
> External email : Please do not click links or open attachments until
> you have verified the sender or the content.
>  On Fri, Jun 09, 2023 at 07:09:00PM +0800, jim.tsai wrote:
> > Export symbols for arch_vma_name and smap_gather_stats
> > functions so that we can detect the memory leak issues.
> > Besides, we can know which memory type is leaked, too.
> 
> This commit description doesn't give enough information. How does
> exporting arch_vma_name() help with detecting memory leak issues?
> 
> You haven't included any users of these new exports, so the initial
> reaction is going to be negative - please include the users of these
> new symbols in your patch set.
> 
> Thanks.
> 
> -- 
> RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
> FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!


Hi Russell,

arch_vma_name() is to get the heap infromation from a user process.

We use these two export functions from our kernel module to get a
specific user process's memory information and heap usage. Furthermore,
we can use such information to detect the memory leak issues. 

The example code is as follows:

/*example code*/

void m_map_vma(struct vm_area_struct *vma, unsigned long cur_pss,
                        unsigned long *native_heap, unsigned long
*java_heap)
{
	struct mm_struct *mm = vma->vm_mm;
        const char *name = NULL;

	...
	name = arch_vma_name(vma);
        if (!name) {
                struct anon_vma_name *anon_name;

                ...
                anon_name = m_anon_vma_name(vma);
                if (anon_name) {
                        if (strstr(anon_name->name, "scudo"))
                                (*native_heap) += cur_pss;
                        else if (strstr(anon_name->name,
"libc_malloc"))
                                (*native_heap) += cur_pss;
                        else if (strstr(anon_name->name, "GWP-ASan"))
                                (*native_heap) += cur_pss;
                        else if (strstr(anon_name->name, "dalvik-alloc
space"))
                                (*java_heap) += cur_pss;
                        else if (strstr(anon_name->name, "dalvik-main
space"))
                                (*java_heap) += cur_pss;

                        ...
                }
        }
}

void calculate_process_memory(struct task_struct *t)
{
        struct mm_struct *mm = NULL;
        struct vm_area_struct *vma = NULL;
        struct mem_size_stats mss;
        unsigned long pss, uss, rss, swap, cur_pss;
        unsigned long java_heap = 0, native_heap = 0;
        struct vma_iterator vmi;

        get_task_struct(t);
        mm = t->mm;
        if (mm) {
                memset(&mss, 0, sizeof(mss));
                mmgrab(mm);
                mmap_read_lock(mm);
                vma_iter_init(&vmi, mm, 0);

                for_each_vma(vmi, vma) {
                        cur_pss = (unsigned long)(mss.pss >>
PSS_SHIFT);
                        smap_gather_stats(vma, &mss, 0);
                        cur_pss =
                                ((unsigned long)(mss.pss >> PSS_SHIFT))
- cur_pss;
                        cur_pss = cur_pss / 1024;
                        m_map_vma(vma, cur_pss, &native_heap,
&java_heap);
                }
                mmap_read_unlock(mm);
                mmdrop(mm);
        }
        put_task_struct(t);

        ...
}

Thanks

Jim