On Thu, 28 Jan 2016 10:26:44 +0100, Vinod Koul wrote:
On Thu, Jan 28, 2016 at 06:49:48AM +0100, Takashi Iwai wrote:
Yes did, but didn't get a clue. tlv seems to be valid before free and points to right location and contents
Could you show the backtrace?
676 if ((err = snd_hctl_elem_tlv_read(elem, tlv, 4096)) < 0) { (gdb) n 677 error("Control %s element TLV read error: %s\n", card, snd_strerror(err)); (gdb) p err $1 = -14
(gdb) n amixer: Control hw:0 element TLV read error: Bad address
678 free(tlv); (gdb) p tlv $2 = (unsigned int *) 0x625f10 (gdb) p tlv[0] $3 = 4294967295 (gdb) p tlv[1] $4 = 0 (gdb) backtrace #0 show_control (elem=0x625310, level=level@entry=5, space=0x409b01 " ") at amixer.c:678 #1 0x00000000004066d9 in cset (argc=argc@entry=1, argv=0x7fffffffe530, roflag=roflag@entry=1, keep_handle=keep_handle@entry=0) at amixer.c:1184 #2 0x0000000000404352 in main (argc=<optimized out>, argv=0x7fffffffe518) at amixer.c:1863 (gdb) c Continuing.
The line number doesn't match with the latest code in git, so double-check that the problem happens with the latest alsa-lib and alsa-utils, too.
I'm thinking whether this is rather an issue in the kernel driver side. In skl_tplg_tlv_control_get(),
if (bc->params) { if (copy_to_user(data, &bc->param_id, sizeof(u32))) return -EFAULT; if (copy_to_user(data + 1, &size, sizeof(u32))) return -EFAULT; if (copy_to_user(data + 2, bc->params, size)) return -EFAULT; }
But here, size is the size of the whole container, not the size in the container. In the code above, you're copying size+8 bytes total and this breaks the boundary already.
Takashi