When swiotlb is full, the kernel would print io_tlb_used. However, the
result might be inaccurate at that time because we have left the critical
section protected by spinlock.
Therefore, we backup the io_tlb_used into local variable before leaving
critical section.
Fixes: 83ca25948940 ("swiotlb: dump used and total slots when swiotlb buffer is full")
Suggested-by: Håkon Bugge <[email protected]>
Signed-off-by: Dongli Zhang <[email protected]>
---
kernel/dma/swiotlb.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/kernel/dma/swiotlb.c b/kernel/dma/swiotlb.c
index 857e823..6f7619c 100644
--- a/kernel/dma/swiotlb.c
+++ b/kernel/dma/swiotlb.c
@@ -452,6 +452,7 @@ phys_addr_t swiotlb_tbl_map_single(struct device *hwdev,
unsigned long mask;
unsigned long offset_slots;
unsigned long max_slots;
+ unsigned long tmp_io_tlb_used;
if (no_iotlb_memory)
panic("Can not allocate SWIOTLB buffer earlier and can't now provide you with the DMA bounce buffer");
@@ -538,10 +539,12 @@ phys_addr_t swiotlb_tbl_map_single(struct device *hwdev,
} while (index != wrap);
not_found:
+ tmp_io_tlb_used = io_tlb_used;
+
spin_unlock_irqrestore(&io_tlb_lock, flags);
if (!(attrs & DMA_ATTR_NO_WARN) && printk_ratelimit())
- dev_warn(hwdev, "swiotlb buffer is full (sz: %zd bytes), total %lu, used %lu\n",
- size, io_tlb_nslabs, io_tlb_used);
+ dev_warn(hwdev, "swiotlb buffer is full (sz: %zd bytes), total %lu (slots), used %lu (slots)\n",
+ size, io_tlb_nslabs, tmp_io_tlb_used);
return DMA_MAPPING_ERROR;
found:
io_tlb_used += nslots;
--
2.7.4
I'm good to have this patch, which helps identify the cause of failure is
fragmentation or it really been used up.
On 4/12/19 4:38 AM, Dongli Zhang wrote:
> When swiotlb is full, the kernel would print io_tlb_used. However, the
> result might be inaccurate at that time because we have left the critical
> section protected by spinlock.
>
> Therefore, we backup the io_tlb_used into local variable before leaving
> critical section.
>
> Fixes: 83ca25948940 ("swiotlb: dump used and total slots when swiotlb buffer is full")
> Suggested-by: Håkon Bugge <[email protected]>
> Signed-off-by: Dongli Zhang <[email protected]>
Reviewed-by: Joe Jin <[email protected]>
Thanks,
Joe
As the patch to be fixed is still in Konrad's own tree, I will send the v2 for
the patch to be fixed, instead of an incremental fix.
Dongli Zhang
On 4/12/19 10:13 PM, Joe Jin wrote:
> I'm good to have this patch, which helps identify the cause of failure is
> fragmentation or it really been used up.
>
> On 4/12/19 4:38 AM, Dongli Zhang wrote:
>> When swiotlb is full, the kernel would print io_tlb_used. However, the
>> result might be inaccurate at that time because we have left the critical
>> section protected by spinlock.
>>
>> Therefore, we backup the io_tlb_used into local variable before leaving
>> critical section.
>>
>> Fixes: 83ca25948940 ("swiotlb: dump used and total slots when swiotlb buffer is full")
>> Suggested-by: Håkon Bugge <[email protected]>
>> Signed-off-by: Dongli Zhang <[email protected]>
>
> Reviewed-by: Joe Jin <[email protected]>
>
> Thanks,
> Joe
>
On 04/15/2019 07:26 PM, Konrad Rzeszutek Wilk wrote:
>
>
> On Mon, Apr 15, 2019, 2:50 AM Dongli Zhang <[email protected]
> <mailto:[email protected]>> wrote:
>
> As the patch to be fixed is still in Konrad's own tree, I will send the v2 for
> the patch to be fixed, instead of an incremental fix.
>
>
> I squashed it in.
Thank you very much!
I saw it is at:
https://git.kernel.org/pub/scm/linux/kernel/git/konrad/swiotlb.git/commit/?h=devel/for-linus-5.1&id=53b29c336830db48ad3dc737f88b8c065b1f0851
Dongli Zhang