Maxim Pugachev wrote:
> Hi,
>
> This patch adds additional informational to ddb's "show malloc"
> command about the largest consumers of memory.
>
> This is my first patch for OpenBSD, so I will be very grateful for any
> comments. Thanks in advance!
>
>
> Index: sys/kern/kern_malloc.c
> ===================================================================
> RCS file: /cvs/src/sys/kern/kern_malloc.c,v
> retrieving revision 1.128
> diff -u -p -r1.128 kern_malloc.c
> --- sys/kern/kern_malloc.c 14 Mar 2015 03:38:50 -0000 1.128
> +++ sys/kern/kern_malloc.c 9 Dec 2015 16:50:57 -0000
> @@ -158,6 +158,8 @@ malloc(size_t size, int type, int flags)
> char *savedtype;
> #endif
> #ifdef KMEMSTATS
> + int i;
> + struct ktopcount *kftop;
> struct kmemstats *ksp = &kmemstats[type];
>
> if (((unsigned long)type) <= 1 || ((unsigned long)type) >= M_LAST)
> @@ -338,6 +340,29 @@ out:
> ksp->ks_calls++;
> if (ksp->ks_memuse > ksp->ks_maxused)
> ksp->ks_maxused = ksp->ks_memuse;
> +
> + void *from = __builtin_return_address(0);
> + from = __builtin_extract_return_addr(from);
> + if (!from)
> + goto topout;
Dropping gcc extensions into the middle of the kernel isn't recommended. Also,
this mixes code and declarations.
> +
> + kftop = &ksp->ks_top[0];
> + for (i = 0; i < KMEMSTATS_TOPCOUNT; ++i)
> + {
> + if (ksp->ks_top[i].kt_from == from) {
> + ksp->ks_top[i].kt_calls++;
> + ksp->ks_top[i].kt_memuse += allocsize;
> + goto topout;
> + }
> + else if (ksp->ks_top[i].kt_memuse < kftop->kt_memuse)
> + kftop = &ksp->ks_top[i];
> + }
> +
> + kftop->kt_calls = 1;
> + kftop->kt_memuse = allocsize;
> + kftop->kt_from = from;
I'm not sure what you're trying to do here. The replacement policy seems to
overwrite the first entry every time. I think you'd have to get pretty lucky
for this to actually track the biggest user.
