Kernel Oops introduced in next-20210222 due to get_max_slots return arg size.
In the function find_slots() variable max_slots is zero when boundary_mask is
0xffffffffffffffff.
[ 0.242119] kernel BUG at ./include/linux/iommu-helper.h:30!
[ 0.247793] Internal error: Oops - BUG: 0 [#1] SMP
[ 0.252595] Modules linked in:
[ 0.255657] CPU: 0 PID: 93 Comm: kworker/0:1 Not tainted 5.11.0-next-20210224+ #25
[ 0.263245] Hardware name: Elba ASIC Board (DT)
[ 0.267784] Workqueue: events_freezable mmc_rescan
[ 0.272592] pstate: 60000085 (nZCv daIf -PAN -UAO -TCO BTYPE=--)
[ 0.278612] pc : swiotlb_tbl_map_single+0x2b0/0x6a0
[ 0.283505] lr : swiotlb_tbl_map_single+0x440/0x6a0
[ 0.288395] sp : ffffffc0122736b0
[ 0.291713] x29: ffffffc0122736b0 x28: ffffffc010e30000
[ 0.297039] x27: 00000000bbf58000 x26: 0000000000000000
[ 0.302364] x25: 0000000000000000 x24: 0000000000000001
[ 0.307689] x23: 0000000000000000 x22: 0000000000000000
[ 0.313013] x21: 0000000000000000 x20: 0000000000000000
[ 0.318338] x19: 0000001241fd4600 x18: ffffffc010d288c8
[ 0.323662] x17: 0000000000000007 x16: 0000000000000001
[ 0.328987] x15: ffffffc092273367 x14: 3a424c54204f4920
[ 0.334311] x13: 6572617774666f73 x12: 20726e2030207865
[ 0.339636] x11: 646e692078787820 x10: 3062653737317830
[ 0.344960] x9 : 2074666968732031 x8 : ffffff977cf82368
[ 0.350285] x7 : 0000000000000001 x6 : c0000000ffffefff
[ 0.355609] x5 : 0000000000017fe8 x4 : 0000000000000000
[ 0.360934] x3 : 0000000000000000 x2 : 18b0d50da009d000
[ 0.366258] x1 : 0000000000000000 x0 : 0000000000000042
[ 0.371583] Call trace:
[ 0.374032] swiotlb_tbl_map_single+0x2b0/0x6a0
[ 0.378573] swiotlb_map+0xa8/0x2b0
Signed-off-by: Brad Larson <[email protected]>
---
kernel/dma/swiotlb.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/dma/swiotlb.c b/kernel/dma/swiotlb.c
index 369e4c3a0f2b..c10e855a03bc 100644
--- a/kernel/dma/swiotlb.c
+++ b/kernel/dma/swiotlb.c
@@ -534,7 +534,7 @@ static int find_slots(struct device *dev, phys_addr_t orig_addr,
unsigned long boundary_mask = dma_get_seg_boundary(dev);
dma_addr_t tbl_dma_addr =
phys_to_dma_unencrypted(dev, io_tlb_start) & boundary_mask;
- unsigned int max_slots = get_max_slots(boundary_mask);
+ unsigned long max_slots = get_max_slots(boundary_mask);
unsigned int iotlb_align_mask =
dma_get_min_align_mask(dev) & ~(IO_TLB_SIZE - 1);
unsigned int nslots = nr_slots(alloc_size), stride;
--
2.17.1
On Fri, Feb 26, 2021 at 12:43:07PM -0800, Brad Larson wrote:
> Kernel Oops introduced in next-20210222 due to get_max_slots return arg size.
> In the function find_slots() variable max_slots is zero when boundary_mask is
> 0xffffffffffffffff.
I am looking at the stable/for-linus-5.12 and what I sent out for
a GIT PULL and I believe this is already squashed in:
531 static int find_slots(struct device *dev, phys_addr_t orig_addr,
532 size_t alloc_size)
533 {
534 unsigned long boundary_mask = dma_get_seg_boundary(dev);
535 dma_addr_t tbl_dma_addr =
536 phys_to_dma_unencrypted(dev, io_tlb_start) & boundary_mask;
537 unsigned long max_slots = get_max_slots(boundary_mask);
Could you double-check please?
>
> [ 0.242119] kernel BUG at ./include/linux/iommu-helper.h:30!
> [ 0.247793] Internal error: Oops - BUG: 0 [#1] SMP
> [ 0.252595] Modules linked in:
> [ 0.255657] CPU: 0 PID: 93 Comm: kworker/0:1 Not tainted 5.11.0-next-20210224+ #25
> [ 0.263245] Hardware name: Elba ASIC Board (DT)
> [ 0.267784] Workqueue: events_freezable mmc_rescan
> [ 0.272592] pstate: 60000085 (nZCv daIf -PAN -UAO -TCO BTYPE=--)
> [ 0.278612] pc : swiotlb_tbl_map_single+0x2b0/0x6a0
> [ 0.283505] lr : swiotlb_tbl_map_single+0x440/0x6a0
> [ 0.288395] sp : ffffffc0122736b0
> [ 0.291713] x29: ffffffc0122736b0 x28: ffffffc010e30000
> [ 0.297039] x27: 00000000bbf58000 x26: 0000000000000000
> [ 0.302364] x25: 0000000000000000 x24: 0000000000000001
> [ 0.307689] x23: 0000000000000000 x22: 0000000000000000
> [ 0.313013] x21: 0000000000000000 x20: 0000000000000000
> [ 0.318338] x19: 0000001241fd4600 x18: ffffffc010d288c8
> [ 0.323662] x17: 0000000000000007 x16: 0000000000000001
> [ 0.328987] x15: ffffffc092273367 x14: 3a424c54204f4920
> [ 0.334311] x13: 6572617774666f73 x12: 20726e2030207865
> [ 0.339636] x11: 646e692078787820 x10: 3062653737317830
> [ 0.344960] x9 : 2074666968732031 x8 : ffffff977cf82368
> [ 0.350285] x7 : 0000000000000001 x6 : c0000000ffffefff
> [ 0.355609] x5 : 0000000000017fe8 x4 : 0000000000000000
> [ 0.360934] x3 : 0000000000000000 x2 : 18b0d50da009d000
> [ 0.366258] x1 : 0000000000000000 x0 : 0000000000000042
> [ 0.371583] Call trace:
> [ 0.374032] swiotlb_tbl_map_single+0x2b0/0x6a0
> [ 0.378573] swiotlb_map+0xa8/0x2b0
>
> Signed-off-by: Brad Larson <[email protected]>
> ---
> kernel/dma/swiotlb.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kernel/dma/swiotlb.c b/kernel/dma/swiotlb.c
> index 369e4c3a0f2b..c10e855a03bc 100644
> --- a/kernel/dma/swiotlb.c
> +++ b/kernel/dma/swiotlb.c
> @@ -534,7 +534,7 @@ static int find_slots(struct device *dev, phys_addr_t orig_addr,
> unsigned long boundary_mask = dma_get_seg_boundary(dev);
> dma_addr_t tbl_dma_addr =
> phys_to_dma_unencrypted(dev, io_tlb_start) & boundary_mask;
> - unsigned int max_slots = get_max_slots(boundary_mask);
> + unsigned long max_slots = get_max_slots(boundary_mask);
> unsigned int iotlb_align_mask =
> dma_get_min_align_mask(dev) & ~(IO_TLB_SIZE - 1);
> unsigned int nslots = nr_slots(alloc_size), stride;
> --
> 2.17.1
>
On Fri, Feb 26, 2021 at 01:18:14PM -0800, Brad Larson wrote:
> On Fri, Feb 26, 2021 at 12:43:07PM -0800, Brad Larson wrote:
> > Kernel Oops introduced in next-20210222 due to get_max_slots return arg
> size.
> > In the function find_slots() variable max_slots is zero when
> boundary_mask is
> > 0xffffffffffffffff.
>
> I am looking at the stable/for-linus-5.12 and what I sent out for
> a GIT PULL and I believe this is already squashed in:
>
> 531 static int find_slots(struct device *dev, phys_addr_t orig_addr,
>
> 532 size_t alloc_size)
>
> 533 {
>
> 534 unsigned long boundary_mask = dma_get_seg_boundary(dev);
>
> 535 dma_addr_t tbl_dma_addr =
>
> 536 phys_to_dma_unencrypted(dev, io_tlb_start) &
> boundary_mask;
> 537 unsigned long max_slots = get_max_slots(boundary_mask);
>
> Could you double-check please?
>
>
> Yes this is squashed in the snippet you are showing. The bug was introduced in
> next-20210222 and is still there when I updated to next-20210226 today.
Duh! It should be fixed now for the next one. Thank you!