* [PATCH] IB/core: fix unmap_sg argument
@ 2016-12-02 13:45 Sebastian Ott
2016-12-13 15:10 ` [PATCH resend] " Sebastian Ott
2016-12-14 19:21 ` [PATCH] " Doug Ledford
0 siblings, 2 replies; 3+ messages in thread
From: Sebastian Ott @ 2016-12-02 13:45 UTC (permalink / raw)
To: Doug Ledford, Sean Hefty, Hal Rosenstock; +Cc: linux-rdma, linux-kernel
Hi,
using dapltest on s390 I ran into the following warning:
[ 20.781709] mlx4_core 0000:00:00.0: DMA-API: device driver frees DMA sg list with different entry count [map count=2] [unmap count=1]
[ 20.781760] ------------[ cut here ]------------
[ 20.781767] WARNING: CPU: 4 PID: 1063 at lib/dma-debug.c:1141 check_unmap+0x658/0xa08
[ 20.781769] Modules linked in: rdma_ucm ib_ucm ib_uverbs rdma_cm configfs ib_cm iw_cm [...]
[ 20.781797] CPU: 4 PID: 1063 Comm: dapltest Tainted: G W 4.9.0-rc7-00039-g43c4f67 #65
[ 20.781799] Hardware name: IBM 2964 N96 704 (LPAR)
[ 20.781801] task: 00000000dd8e4008 task.stack: 00000000f1448000
[ 20.781803] Krnl PSW : 0404c00180000000 000000000060af08 (check_unmap+0x658/0xa08)
[ 20.781806] R:0 T:1 IO:0 EX:0 Key:0 M:1 W:0 P:0 AS:3 CC:0 PM:0 RI:0 EA:3
[ 20.781809] Krnl GPRS: 0000000000000002 00000000dd8e4008 0000000000000079 0000000000a1d4f0
[ 20.781811] 000000000060af04 0000000000000000 0000000000000001 0000000000000001
[ 20.781813] 00000000f8e8e880 07000000008f2210 0000000001e47700 00000000f7a11298
[ 20.781814] 00000000f144ba68 00000000008f21f8 000000000060af04 00000000f144b960
[ 20.781822] Krnl Code: 000000000060aef8: c0200022fcc0 larl %r2,a6a878
000000000060aefe: c0e5ffe474dd brasl %r14,2998b8
#000000000060af04: a7f40001 brc 15,60af06
>000000000060af08: c0200022f976 larl %r2,a6a1f4
000000000060af0e: c0e5ffe474d5 brasl %r14,2998b8
000000000060af14: 4120b050 la %r2,80(%r11)
000000000060af18: a7390000 lghi %r3,0
000000000060af1c: c0e5ffde780a brasl %r14,1d9f30
[ 20.781847] Call Trace:
[ 20.781848] ([<000000000060af04>] check_unmap+0x654/0xa08)
[ 20.781851] [<000000000060b6bc>] debug_dma_unmap_sg+0xf4/0x160
[ 20.781873] [<000003ff82245738>] __ib_umem_release+0x98/0x1a8 [ib_core]
[ 20.781881] [<000003ff82245f3e>] ib_umem_release+0x5e/0x1d0 [ib_core]
[ 20.781891] [<000003ff80139976>] mlx4_ib_destroy_qp+0x47e/0x550 [mlx4_ib]
[ 20.781898] [<000003ff8222ad2e>] ib_destroy_qp+0x116/0x188 [ib_core]
[ 20.781902] [<000003ff80052fba>] ib_uverbs_destroy_qp+0xb2/0x1a0 [ib_uverbs]
[ 20.781905] [<000003ff8004cada>] ib_uverbs_write+0x20a/0x488 [ib_uverbs]
[ 20.781910] [<0000000000352756>] __vfs_write+0x36/0x138
[ 20.781912] [<000000000035348c>] vfs_write+0xbc/0x1a0
[ 20.781914] [<0000000000354a96>] SyS_write+0x66/0xc0
[ 20.781918] [<000000000087d796>] system_call+0xd6/0x270
[ 20.781919] INFO: lockdep is turned off.
[ 20.781921] Last Breaking-Event-Address:
[ 20.781922] [<000000000060af04>] check_unmap+0x654/0xa08
[ 20.781924] ---[ end trace bd581c43b9bebeea ]---
[ 20.781925] Mapped at:
[ 20.781929] [<000000000060b4a6>] debug_dma_map_sg+0x5e/0x180
[ 20.781938] [<000003ff82245d12>] ib_umem_get+0x4ca/0x610 [ib_core]
[ 20.781943] [<000003ff80136df2>] create_qp_common.isra.15+0x572/0x1010 [mlx4_ib]
[ 20.781949] [<000003ff8013929e>] mlx4_ib_create_qp+0x1de/0x438 [mlx4_ib]
[ 20.781953] [<000003ff8004f52c>] create_qp.isra.11+0x44c/0x7f0 [ib_uverbs]
The following patch fixes this:
---->8
>From ec91646d8c14e2a8dd2b62187084dab32ef8a56b Mon Sep 17 00:00:00 2001
From: Sebastian Ott <sebott@linux.vnet.ibm.com>
Date: Fri, 2 Dec 2016 11:15:05 +0100
Subject: [PATCH] IB/core: fix unmap_sg argument
__ib_umem_release calls dma_unmap_sg with a different number of
sg_entries than ib_umem_get uses for dma_map_sg. This might cause
trouble for implementations that merge sglist entries and results
in the following dma debug complaint:
DMA-API: device driver frees DMA sg list with different entry
count [map count=2] [unmap count=1]
Fix it by using the correct value.
Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
---
drivers/infiniband/core/umem.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c
index 84b4eff..1e62a5f 100644
--- a/drivers/infiniband/core/umem.c
+++ b/drivers/infiniband/core/umem.c
@@ -51,7 +51,7 @@ static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int d
if (umem->nmap > 0)
ib_dma_unmap_sg(dev, umem->sg_head.sgl,
- umem->nmap,
+ umem->npages,
DMA_BIDIRECTIONAL);
for_each_sg(umem->sg_head.sgl, sg, umem->npages, i) {
--
2.7.4
^ permalink raw reply related [flat|nested] 3+ messages in thread
* [PATCH resend] IB/core: fix unmap_sg argument
2016-12-02 13:45 [PATCH] IB/core: fix unmap_sg argument Sebastian Ott
@ 2016-12-13 15:10 ` Sebastian Ott
2016-12-14 19:21 ` [PATCH] " Doug Ledford
1 sibling, 0 replies; 3+ messages in thread
From: Sebastian Ott @ 2016-12-13 15:10 UTC (permalink / raw)
To: Doug Ledford, Sean Hefty, Hal Rosenstock; +Cc: linux-rdma, linux-kernel
Hi,
using dapltest on s390 I ran into the following warning:
[ 20.781709] mlx4_core 0000:00:00.0: DMA-API: device driver frees DMA sg list with different entry count [map count=2] [unmap count=1]
[ 20.781760] ------------[ cut here ]------------
[ 20.781767] WARNING: CPU: 4 PID: 1063 at lib/dma-debug.c:1141 check_unmap+0x658/0xa08
[ 20.781769] Modules linked in: rdma_ucm ib_ucm ib_uverbs rdma_cm configfs ib_cm iw_cm [...]
[ 20.781797] CPU: 4 PID: 1063 Comm: dapltest Tainted: G W 4.9.0-rc7-00039-g43c4f67 #65
[ 20.781799] Hardware name: IBM 2964 N96 704 (LPAR)
[ 20.781801] task: 00000000dd8e4008 task.stack: 00000000f1448000
[ 20.781803] Krnl PSW : 0404c00180000000 000000000060af08 (check_unmap+0x658/0xa08)
[ 20.781806] R:0 T:1 IO:0 EX:0 Key:0 M:1 W:0 P:0 AS:3 CC:0 PM:0 RI:0 EA:3
[ 20.781809] Krnl GPRS: 0000000000000002 00000000dd8e4008 0000000000000079 0000000000a1d4f0
[ 20.781811] 000000000060af04 0000000000000000 0000000000000001 0000000000000001
[ 20.781813] 00000000f8e8e880 07000000008f2210 0000000001e47700 00000000f7a11298
[ 20.781814] 00000000f144ba68 00000000008f21f8 000000000060af04 00000000f144b960
[ 20.781822] Krnl Code: 000000000060aef8: c0200022fcc0 larl %r2,a6a878
000000000060aefe: c0e5ffe474dd brasl %r14,2998b8
#000000000060af04: a7f40001 brc 15,60af06
>000000000060af08: c0200022f976 larl %r2,a6a1f4
000000000060af0e: c0e5ffe474d5 brasl %r14,2998b8
000000000060af14: 4120b050 la %r2,80(%r11)
000000000060af18: a7390000 lghi %r3,0
000000000060af1c: c0e5ffde780a brasl %r14,1d9f30
[ 20.781847] Call Trace:
[ 20.781848] ([<000000000060af04>] check_unmap+0x654/0xa08)
[ 20.781851] [<000000000060b6bc>] debug_dma_unmap_sg+0xf4/0x160
[ 20.781873] [<000003ff82245738>] __ib_umem_release+0x98/0x1a8 [ib_core]
[ 20.781881] [<000003ff82245f3e>] ib_umem_release+0x5e/0x1d0 [ib_core]
[ 20.781891] [<000003ff80139976>] mlx4_ib_destroy_qp+0x47e/0x550 [mlx4_ib]
[ 20.781898] [<000003ff8222ad2e>] ib_destroy_qp+0x116/0x188 [ib_core]
[ 20.781902] [<000003ff80052fba>] ib_uverbs_destroy_qp+0xb2/0x1a0 [ib_uverbs]
[ 20.781905] [<000003ff8004cada>] ib_uverbs_write+0x20a/0x488 [ib_uverbs]
[ 20.781910] [<0000000000352756>] __vfs_write+0x36/0x138
[ 20.781912] [<000000000035348c>] vfs_write+0xbc/0x1a0
[ 20.781914] [<0000000000354a96>] SyS_write+0x66/0xc0
[ 20.781918] [<000000000087d796>] system_call+0xd6/0x270
[ 20.781919] INFO: lockdep is turned off.
[ 20.781921] Last Breaking-Event-Address:
[ 20.781922] [<000000000060af04>] check_unmap+0x654/0xa08
[ 20.781924] ---[ end trace bd581c43b9bebeea ]---
[ 20.781925] Mapped at:
[ 20.781929] [<000000000060b4a6>] debug_dma_map_sg+0x5e/0x180
[ 20.781938] [<000003ff82245d12>] ib_umem_get+0x4ca/0x610 [ib_core]
[ 20.781943] [<000003ff80136df2>] create_qp_common.isra.15+0x572/0x1010 [mlx4_ib]
[ 20.781949] [<000003ff8013929e>] mlx4_ib_create_qp+0x1de/0x438 [mlx4_ib]
[ 20.781953] [<000003ff8004f52c>] create_qp.isra.11+0x44c/0x7f0 [ib_uverbs]
The following patch fixes this:
---->8
>From ec91646d8c14e2a8dd2b62187084dab32ef8a56b Mon Sep 17 00:00:00 2001
From: Sebastian Ott <sebott@linux.vnet.ibm.com>
Date: Fri, 2 Dec 2016 11:15:05 +0100
Subject: [PATCH] IB/core: fix unmap_sg argument
__ib_umem_release calls dma_unmap_sg with a different number of
sg_entries than ib_umem_get uses for dma_map_sg. This might cause
trouble for implementations that merge sglist entries and results
in the following dma debug complaint:
DMA-API: device driver frees DMA sg list with different entry
count [map count=2] [unmap count=1]
Fix it by using the correct value.
Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
---
drivers/infiniband/core/umem.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c
index 84b4eff..1e62a5f 100644
--- a/drivers/infiniband/core/umem.c
+++ b/drivers/infiniband/core/umem.c
@@ -51,7 +51,7 @@ static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int d
if (umem->nmap > 0)
ib_dma_unmap_sg(dev, umem->sg_head.sgl,
- umem->nmap,
+ umem->npages,
DMA_BIDIRECTIONAL);
for_each_sg(umem->sg_head.sgl, sg, umem->npages, i) {
--
2.7.4
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] IB/core: fix unmap_sg argument
2016-12-02 13:45 [PATCH] IB/core: fix unmap_sg argument Sebastian Ott
2016-12-13 15:10 ` [PATCH resend] " Sebastian Ott
@ 2016-12-14 19:21 ` Doug Ledford
1 sibling, 0 replies; 3+ messages in thread
From: Doug Ledford @ 2016-12-14 19:21 UTC (permalink / raw)
To: Sebastian Ott, Sean Hefty, Hal Rosenstock; +Cc: linux-rdma, linux-kernel
[-- Attachment #1.1: Type: text/plain, Size: 1218 bytes --]
On 12/2/2016 8:45 AM, Sebastian Ott wrote:
> __ib_umem_release calls dma_unmap_sg with a different number of
> sg_entries than ib_umem_get uses for dma_map_sg. This might cause
> trouble for implementations that merge sglist entries and results
> in the following dma debug complaint:
>
> DMA-API: device driver frees DMA sg list with different entry
> count [map count=2] [unmap count=1]
>
> Fix it by using the correct value.
>
> Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
> ---
> drivers/infiniband/core/umem.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c
> index 84b4eff..1e62a5f 100644
> --- a/drivers/infiniband/core/umem.c
> +++ b/drivers/infiniband/core/umem.c
> @@ -51,7 +51,7 @@ static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int d
>
> if (umem->nmap > 0)
> ib_dma_unmap_sg(dev, umem->sg_head.sgl,
> - umem->nmap,
> + umem->npages,
> DMA_BIDIRECTIONAL);
>
> for_each_sg(umem->sg_head.sgl, sg, umem->npages, i) {
>
Thanks, applied.
--
Doug Ledford <dledford@redhat.com>
GPG Key ID: 0E572FDD
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 884 bytes --]
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2016-12-14 19:22 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-12-02 13:45 [PATCH] IB/core: fix unmap_sg argument Sebastian Ott
2016-12-13 15:10 ` [PATCH resend] " Sebastian Ott
2016-12-14 19:21 ` [PATCH] " Doug Ledford
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).