While testing for leaks in an unrelated part of the kernel, my ::findleaks only discovered leaks in the LX brand code in the kernel:
> ::findleaks -d
CACHE LEAKED BUFCTL CALLER
fffffeb187cd1008 17 fffffebb99bd8558 lx_mount+0x10e
fffffeb187cd1008 18 fffffeb45fb3c390 lx_mount+0x121
fffffeb187cd1008 17 fffffebb9b8e3e90 lx_mount+0x130
fffffeb187cd1008 2 fffffebc262e63c8 pn_get+0x35
------------------------------------------------------------------------
Total 54 buffers, 62208 bytes
kmem_alloc_1152 leak: 17 buffers, 1152 bytes each, 19584 bytes total
ADDR BUFADDR TIMESTAMP THREAD
CACHE LASTLOG CONTENTS
fffffebb99bd8558 fffffebb9b6ae180 12599210e2a1b fffffebad3633be0
fffffeb187cd1008 fffffeb1cb9c3f00 fffffeb2800e8788
kmem_cache_alloc_debug+0x37c
kmem_cache_alloc+0x138
kmem_alloc+0x4a
lx_mount+0x10e
lx_syscall_enter+0x1b2
kmem_alloc_1152 leak: 18 buffers, 1152 bytes each, 20736 bytes total
ADDR BUFADDR TIMESTAMP THREAD
CACHE LASTLOG CONTENTS
fffffeb45fb3c390 fffffebba45f0000 12599210e2b77 fffffebad3633be0
fffffeb187cd1008 fffffeb1cb9c3fc0 fffffeb2800e8688
kmem_cache_alloc_debug+0x37c
kmem_cache_alloc+0x138
kmem_alloc+0x4a
lx_mount+0x121
lx_syscall_enter+0x1b2
kmem_alloc_1152 leak: 17 buffers, 1152 bytes each, 19584 bytes total
ADDR BUFADDR TIMESTAMP THREAD
CACHE LASTLOG CONTENTS
fffffebb9b8e3e90 fffffebb9dfaf540 12599210e2ccd fffffebad3633be0
fffffeb187cd1008 fffffeb1cb9c4080 fffffeb2800e8588
kmem_cache_alloc_debug+0x37c
kmem_cache_alloc+0x138
kmem_alloc+0x4a
lx_mount+0x130
lx_syscall_enter+0x1b2
kmem_alloc_1152 leak: 2 buffers, 1152 bytes each, 2304 bytes total
ADDR BUFADDR TIMESTAMP THREAD
CACHE LASTLOG CONTENTS
fffffebc262e63c8 fffffebc14f48000 1259d3ae6c208 fffffeb43cc1d100
fffffeb187cd1008 fffffeb1c18fd580 fffffeb272d18348
kmem_cache_alloc_debug+0x37c
kmem_cache_alloc+0x138
kmem_alloc+0x4a
pn_get+0x35
tmp_mount+0x1d2
fsop_mount+0x24
domount+0x47e
lx_mount+0x4f4
lx_syscall_enter+0x1b2
> An inspection of lx_mount(), the function all of these leaks have in common, did not show an OBVIOUS leak, as the out label kmem_frees what it allocated.. It is possible this is system-shutdown-time leaks (found in a reboot -d dump), but this should be investigated in a controlled environment, or we need to free things if the domount() call fails..