Re: [BUG] net: sungem: device driver frees DMA memory with wrong function
On Fri, Dec 28, 2018 at 12:36:21AM -0800, Christoph Hellwig wrote:
> Please try this patch:
>
The error type change to "DMA-API: gem :00:0f.0: device driver failed to
check map error" (I will send patch for fixing this).
Note that I used the patch from your just sent DMA series (since the patch
below is included in).
So you can add my
Tested-by: LABBE Corentin
Thanks
> diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
> index a52c6409bdc2..f454e0ed1398 100644
> --- a/include/linux/dma-mapping.h
> +++ b/include/linux/dma-mapping.h
> @@ -284,32 +284,25 @@ static inline void dma_direct_sync_sg_for_cpu(struct
> device *dev,
> }
> #endif
>
> -static inline dma_addr_t dma_map_single_attrs(struct device *dev, void *ptr,
> - size_t size,
> - enum dma_data_direction dir,
> - unsigned long attrs)
> +static inline dma_addr_t dma_map_page_attrs(struct device *dev,
> + struct page *page, size_t offset, size_t size,
> + enum dma_data_direction dir, unsigned long attrs)
> {
> const struct dma_map_ops *ops = get_dma_ops(dev);
> dma_addr_t addr;
>
> BUG_ON(!valid_dma_direction(dir));
> - debug_dma_map_single(dev, ptr, size);
> if (dma_is_direct(ops))
> - addr = dma_direct_map_page(dev, virt_to_page(ptr),
> - offset_in_page(ptr), size, dir, attrs);
> + addr = dma_direct_map_page(dev, page, offset, size, dir, attrs);
> else
> - addr = ops->map_page(dev, virt_to_page(ptr),
> - offset_in_page(ptr), size, dir, attrs);
> - debug_dma_map_page(dev, virt_to_page(ptr),
> -offset_in_page(ptr), size,
> -dir, addr, true);
> + addr = ops->map_page(dev, page, offset, size, dir, attrs);
> + debug_dma_map_page(dev, page, offset, size, dir, addr, false);
> +
> return addr;
> }
>
> -static inline void dma_unmap_single_attrs(struct device *dev, dma_addr_t
> addr,
> - size_t size,
> - enum dma_data_direction dir,
> - unsigned long attrs)
> +static inline void dma_unmap_page_attrs(struct device *dev, dma_addr_t addr,
> + size_t size, enum dma_data_direction dir, unsigned long attrs)
> {
> const struct dma_map_ops *ops = get_dma_ops(dev);
>
> @@ -321,12 +314,6 @@ static inline void dma_unmap_single_attrs(struct device
> *dev, dma_addr_t addr,
> debug_dma_unmap_page(dev, addr, size, dir, true);
> }
>
> -static inline void dma_unmap_page_attrs(struct device *dev, dma_addr_t addr,
> - size_t size, enum dma_data_direction dir, unsigned long attrs)
> -{
> - return dma_unmap_single_attrs(dev, addr, size, dir, attrs);
> -}
> -
> /*
> * dma_maps_sg_attrs returns 0 on error and > 0 on success.
> * It should never return a value < 0.
> @@ -363,25 +350,6 @@ static inline void dma_unmap_sg_attrs(struct device
> *dev, struct scatterlist *sg
> ops->unmap_sg(dev, sg, nents, dir, attrs);
> }
>
> -static inline dma_addr_t dma_map_page_attrs(struct device *dev,
> - struct page *page,
> - size_t offset, size_t size,
> - enum dma_data_direction dir,
> - unsigned long attrs)
> -{
> - const struct dma_map_ops *ops = get_dma_ops(dev);
> - dma_addr_t addr;
> -
> - BUG_ON(!valid_dma_direction(dir));
> - if (dma_is_direct(ops))
> - addr = dma_direct_map_page(dev, page, offset, size, dir, attrs);
> - else
> - addr = ops->map_page(dev, page, offset, size, dir, attrs);
> - debug_dma_map_page(dev, page, offset, size, dir, addr, false);
> -
> - return addr;
> -}
> -
> static inline dma_addr_t dma_map_resource(struct device *dev,
> phys_addr_t phys_addr,
> size_t size,
> @@ -488,6 +456,19 @@ dma_sync_sg_for_device(struct device *dev, struct
> scatterlist *sg,
>
> }
>
> +static inline dma_addr_t dma_map_single_attrs(struct device *dev, void *ptr,
> + size_t size, enum dma_data_direction dir, unsigned long attrs)
> +{
> + return dma_map_page_attrs(dev, virt_to_page(ptr), offset_in_page(ptr),
> + size, dir, attrs);
> +}
> +
> +static inline void dma_unmap_single_attrs(struct device *dev, dma_addr_t
> addr,
> + size_t size, enum dma_data_direction dir, unsigned long attrs)
> +{
> + return dma_unmap_page_attrs(dev, addr, size, dir, attrs);
> +}
> +
> #define dma_map_single(d, a, s, r) dma_map_single_attrs(d, a, s, r, 0)
> #define dma_unmap_single(d, a, s, r)
Re: [BUG] net: sungem: device driver frees DMA memory with wrong function
Please try this patch:
diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
index a52c6409bdc2..f454e0ed1398 100644
--- a/include/linux/dma-mapping.h
+++ b/include/linux/dma-mapping.h
@@ -284,32 +284,25 @@ static inline void dma_direct_sync_sg_for_cpu(struct
device *dev,
}
#endif
-static inline dma_addr_t dma_map_single_attrs(struct device *dev, void *ptr,
- size_t size,
- enum dma_data_direction dir,
- unsigned long attrs)
+static inline dma_addr_t dma_map_page_attrs(struct device *dev,
+ struct page *page, size_t offset, size_t size,
+ enum dma_data_direction dir, unsigned long attrs)
{
const struct dma_map_ops *ops = get_dma_ops(dev);
dma_addr_t addr;
BUG_ON(!valid_dma_direction(dir));
- debug_dma_map_single(dev, ptr, size);
if (dma_is_direct(ops))
- addr = dma_direct_map_page(dev, virt_to_page(ptr),
- offset_in_page(ptr), size, dir, attrs);
+ addr = dma_direct_map_page(dev, page, offset, size, dir, attrs);
else
- addr = ops->map_page(dev, virt_to_page(ptr),
- offset_in_page(ptr), size, dir, attrs);
- debug_dma_map_page(dev, virt_to_page(ptr),
- offset_in_page(ptr), size,
- dir, addr, true);
+ addr = ops->map_page(dev, page, offset, size, dir, attrs);
+ debug_dma_map_page(dev, page, offset, size, dir, addr, false);
+
return addr;
}
-static inline void dma_unmap_single_attrs(struct device *dev, dma_addr_t addr,
- size_t size,
- enum dma_data_direction dir,
- unsigned long attrs)
+static inline void dma_unmap_page_attrs(struct device *dev, dma_addr_t addr,
+ size_t size, enum dma_data_direction dir, unsigned long attrs)
{
const struct dma_map_ops *ops = get_dma_ops(dev);
@@ -321,12 +314,6 @@ static inline void dma_unmap_single_attrs(struct device
*dev, dma_addr_t addr,
debug_dma_unmap_page(dev, addr, size, dir, true);
}
-static inline void dma_unmap_page_attrs(struct device *dev, dma_addr_t addr,
- size_t size, enum dma_data_direction dir, unsigned long attrs)
-{
- return dma_unmap_single_attrs(dev, addr, size, dir, attrs);
-}
-
/*
* dma_maps_sg_attrs returns 0 on error and > 0 on success.
* It should never return a value < 0.
@@ -363,25 +350,6 @@ static inline void dma_unmap_sg_attrs(struct device *dev,
struct scatterlist *sg
ops->unmap_sg(dev, sg, nents, dir, attrs);
}
-static inline dma_addr_t dma_map_page_attrs(struct device *dev,
- struct page *page,
- size_t offset, size_t size,
- enum dma_data_direction dir,
- unsigned long attrs)
-{
- const struct dma_map_ops *ops = get_dma_ops(dev);
- dma_addr_t addr;
-
- BUG_ON(!valid_dma_direction(dir));
- if (dma_is_direct(ops))
- addr = dma_direct_map_page(dev, page, offset, size, dir, attrs);
- else
- addr = ops->map_page(dev, page, offset, size, dir, attrs);
- debug_dma_map_page(dev, page, offset, size, dir, addr, false);
-
- return addr;
-}
-
static inline dma_addr_t dma_map_resource(struct device *dev,
phys_addr_t phys_addr,
size_t size,
@@ -488,6 +456,19 @@ dma_sync_sg_for_device(struct device *dev, struct
scatterlist *sg,
}
+static inline dma_addr_t dma_map_single_attrs(struct device *dev, void *ptr,
+ size_t size, enum dma_data_direction dir, unsigned long attrs)
+{
+ return dma_map_page_attrs(dev, virt_to_page(ptr), offset_in_page(ptr),
+ size, dir, attrs);
+}
+
+static inline void dma_unmap_single_attrs(struct device *dev, dma_addr_t addr,
+ size_t size, enum dma_data_direction dir, unsigned long attrs)
+{
+ return dma_unmap_page_attrs(dev, addr, size, dir, attrs);
+}
+
#define dma_map_single(d, a, s, r) dma_map_single_attrs(d, a, s, r, 0)
#define dma_unmap_single(d, a, s, r) dma_unmap_single_attrs(d, a, s, r, 0)
#define dma_map_sg(d, s, n, r) dma_map_sg_attrs(d, s, n, r, 0)
Re: [BUG] net: sungem: device driver frees DMA memory with wrong function
From: Corentin Labbe Date: Sun, 23 Dec 2018 20:38:21 +0100 > During a boot on a qemu machine, I hit the following problem: ... > After some pr_info, I found that the function triggering this is the > pci_unmap_page() in gem_tx(). > So clearly this WARNING() is strange since it says "unmapped as single". I also went through the code paths and this makes no sense to me either.
[BUG] net: sungem: device driver frees DMA memory with wrong function
Hello During a boot on a qemu machine, I hit the following problem: [ 21.613659] [ cut here ] [ 21.614039] DMA-API: gem :00:0f.0: device driver frees DMA memory with wrong function [device address=0x185c5402] [size=408 bytes] [mapped as page] [unmapped as single] [ 21.615960] WARNING: CPU: 0 PID: 206 at /linux-next/kernel/dma/debug.c:1085 check_unmap+0x1b0/0xd10 [ 21.616599] Modules linked in: [ 21.617344] CPU: 0 PID: 206 Comm: dhcpcd Not tainted 4.20.0-rc6-next-20181217+ #12 [ 21.617735] NIP: c00ad4e0 LR: c00ad4e0 CTR: c058a710 [ 21.618068] REGS: dfff7cb0 TRAP: 0700 Not tainted (4.20.0-rc6-next-20181217+) [ 21.618404] MSR: 00021032 CR: 28008264 XER: [ 21.618789] [ 21.618789] GPR00: c00ad4e0 dfff7d60 d85924c0 00a8 0102 0101 00fe 5d205b6d [ 21.618789] GPR08: 0007 00fe 22008864 100581b4 dfff7f1c de019508 [ 21.618789] GPR16: de019000 0001 c0a67f00 0004 c0b44018 c0a31fc4 c0b43bdc [ 21.618789] GPR24: 9032 c0dadedc c0db c0da999c c0b43bdc c0c0b178 dfff7de0 de076ce8 [ 21.620219] NIP [c00ad4e0] check_unmap+0x1b0/0xd10 [ 21.620478] LR [c00ad4e0] check_unmap+0x1b0/0xd10 [ 21.620703] Call Trace: [ 21.620963] [dfff7d60] [c00ad4e0] check_unmap+0x1b0/0xd10 (unreliable) [ 21.621379] [dfff7dd0] [c00ae128] debug_dma_unmap_page+0xe8/0x120 [ 21.621656] [dfff7e40] [c05a865c] gem_poll+0x1000/0x18fc [ 21.621863] [dfff7f00] [c071f044] net_rx_action+0x1b8/0x41c [ 21.622242] [dfff7f80] [c08a287c] __do_softirq+0x17c/0x3b0 [ 21.622495] [dfff7ff0] [c0011378] call_do_softirq+0x24/0x3c [ 21.622697] [d86c9c20] [c000724c] do_softirq_own_stack+0x3c/0x7c [ 21.623008] [d86c9c40] [c004e800] do_softirq.part.1+0x64/0x7c [ 21.623292] [d86c9c60] [c004e8d4] __local_bh_enable_ip+0xbc/0x138 [ 21.623526] [d86c9c80] [c071c9b4] __dev_queue_xmit+0x28c/0x874 [ 21.623776] [d86c9cf0] [c083ebe8] packet_snd+0x4f8/0x988 [ 21.624057] [d86c9d80] [c06eddb8] sock_sendmsg+0x20/0x40 [ 21.624410] [d86c9d90] [c06ede94] sock_write_iter+0xbc/0x138 [ 21.624636] [d86c9df0] [c018f50c] do_iter_readv_writev+0x1dc/0x1f4 [ 21.624872] [d86c9e40] [c01925a8] do_iter_write+0xb4/0x350 [ 21.625143] [d86c9e80] [c01928f4] vfs_writev+0x88/0x140 [ 21.625446] [d86c9f00] [c0192a1c] do_writev+0x70/0x104 [ 21.625621] [d86c9f40] [c001912c] ret_from_syscall+0x0/0x38 [ 21.626001] --- interrupt: c01 at 0xfedbd08 [ 21.626001] LR = 0xffbc1f8 [ 21.626357] Instruction dump: [ 21.626719] 7eb5ba14 7e95a214 2b940014 419d0970 92c10008 7efcba14 3c60c0a1 7e649b78 [ 21.627152] 38637cac 80d7043c 90c1000c 4bf9d36d <0fe0> 8261003c 82810040 82a10044 [ 21.627665] ---[ end trace fd53714bd2a5fd06 ]--- After some pr_info, I found that the function triggering this is the pci_unmap_page() in gem_tx(). So clearly this WARNING() is strange since it says "unmapped as single". The qemu is ran with: qemu-system-ppc -machine mac99,via=pmu -nographic -net nic,model=sungem,macaddr=52:54:00:12:34:58 -net tap -m 512 -monitor none -append "console=ttyPZ0 root=/dev/ram0" -kernel vmlinux -initrd rootfs.cpio.gz -drive format=qcow2,file=lava-guest.qcow2,media=disk,id=test,if=ide Regards
