Hi Konrad,
This patch series adds a kernel command line option to aid debugging
when developing support for DMA to memory outside the 32-bit address
space. If specified, trying to map memory that cannot be used with DMA
will fail, and a warning will be printed. This has been used
succesfully to find drivers and DMA engines that do not support 64-bit
memory.
Changes compared to v1:
- Dropp patch "swiotlb: Rate-limit printing when running out of
SW-IOMMU space", which has been picked up,
- Add cleanup patch 1/3,
- Add patch 2/3, to convert the flag swiotlb_force from an int to an
enum,
- Change the kernel parameter "swiotlb=nobounce" to "swiotlb=noforce"
(requested by Konrad Wilk),
- Extend swiotlb_force enum instead of adding an swiotlb_nobounce
variable (requested by Konrad Wilk).
This has been tested on ra7796/salvator-x, and compile-tested on x86.
Thanks!
Geert Uytterhoeven (3):
x86, swiotlb: Simplify pci_swiotlb_detect_override()
swiotlb: Convert swiotlb_force from int to enum
swiotlb: Add swiotlb=noforce debug option
Documentation/admin-guide/kernel-parameters.txt | 3 ++-
arch/arm64/mm/dma-mapping.c | 3 ++-
arch/arm64/mm/init.c | 3 ++-
arch/x86/kernel/pci-swiotlb.c | 6 ++----
arch/x86/xen/pci-swiotlb-xen.c | 2 +-
drivers/xen/swiotlb-xen.c | 4 ++--
include/linux/swiotlb.h | 8 +++++++-
include/trace/events/swiotlb.h | 17 +++++++++-------
lib/swiotlb.c | 26 +++++++++++++++++++------
9 files changed, 48 insertions(+), 24 deletions(-)
--
1.9.1
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
On architectures like arm64, swiotlb is tied intimately to the core
architecture DMA support. In addition, ZONE_DMA cannot be disabled.
To aid debugging and catch devices not supporting DMA to memory outside
the 32-bit address space, add a kernel command line option
"swiotlb=noforce", which disables the use of bounce buffers.
If specified, trying to map memory that cannot be used with DMA will
fail, and a rate-limited warning will be printed.
Note that io_tlb_nslabs is set to 1, which is the minimal supported
value.
Signed-off-by: Geert Uytterhoeven <[email protected]>
---
v2:
- Change kernel parameter value from "nobounce" to "noforce",
- Extend swiotlb_force enum instead of adding swiotlb_nobounce
variable (both requested by Konrad Wilk).
---
Documentation/admin-guide/kernel-parameters.txt | 3 ++-
include/linux/swiotlb.h | 1 +
include/trace/events/swiotlb.h | 3 ++-
lib/swiotlb.c | 18 ++++++++++++++++--
4 files changed, 21 insertions(+), 4 deletions(-)
diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 21e2d88637050b7a..24fb691b2da5f78d 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -3811,10 +3811,11 @@
it if 0 is given (See Documentation/cgroup-v1/memory.txt)
swiotlb= [ARM,IA-64,PPC,MIPS,X86]
- Format: { <int> | force }
+ Format: { <int> | force | noforce }
<int> -- Number of I/O TLB slabs
force -- force using of bounce buffers even if they
wouldn't be automatically used by the kernel
+ noforce -- Never use bounce buffers (for debugging)
switches= [HW,M68k]
diff --git a/include/linux/swiotlb.h b/include/linux/swiotlb.h
index 71d104e4c8490763..d9c84a9cde3dd7d7 100644
--- a/include/linux/swiotlb.h
+++ b/include/linux/swiotlb.h
@@ -12,6 +12,7 @@
enum swiotlb_force {
SWIOTLB_NORMAL, /* Default - depending on HW DMA mask etc. */
SWIOTLB_FORCE, /* swiotlb=force */
+ SWIOTLB_NO_FORCE, /* swiotlb=noforce */
};
extern enum swiotlb_force swiotlb_force;
diff --git a/include/trace/events/swiotlb.h b/include/trace/events/swiotlb.h
index 5e2e30a7efce0ebd..288c0c54a2b4ace6 100644
--- a/include/trace/events/swiotlb.h
+++ b/include/trace/events/swiotlb.h
@@ -39,7 +39,8 @@
__entry->size,
__print_symbolic(__entry->swiotlb_force,
{ SWIOTLB_NORMAL, "NORMAL" },
- { SWIOTLB_FORCE, "FORCE" }))
+ { SWIOTLB_FORCE, "FORCE" },
+ { SWIOTLB_NO_FORCE, "NO_FORCE" }))
);
#endif /* _TRACE_SWIOTLB_H */
diff --git a/lib/swiotlb.c b/lib/swiotlb.c
index a32dce6d5101f408..9def738af4f4e568 100644
--- a/lib/swiotlb.c
+++ b/lib/swiotlb.c
@@ -106,8 +106,12 @@
}
if (*str == ',')
++str;
- if (!strcmp(str, "force"))
+ if (!strcmp(str, "force")) {
swiotlb_force = SWIOTLB_FORCE;
+ } else if (!strcmp(str, "noforce")) {
+ swiotlb_force = SWIOTLB_NO_FORCE;
+ io_tlb_nslabs = 1;
+ }
return 0;
}
@@ -543,8 +547,15 @@ phys_addr_t swiotlb_tbl_map_single(struct device *hwdev,
map_single(struct device *hwdev, phys_addr_t phys, size_t size,
enum dma_data_direction dir, unsigned long attrs)
{
- dma_addr_t start_dma_addr = phys_to_dma(hwdev, io_tlb_start);
+ dma_addr_t start_dma_addr;
+
+ if (swiotlb_force == SWIOTLB_NO_FORCE) {
+ dev_warn_ratelimited(hwdev, "Cannot do DMA to address %pa\n",
+ &phys);
+ return SWIOTLB_MAP_ERROR;
+ }
+ start_dma_addr = phys_to_dma(hwdev, io_tlb_start);
return swiotlb_tbl_map_single(hwdev, start_dma_addr, phys, size,
dir, attrs);
}
@@ -721,6 +732,9 @@ void swiotlb_tbl_sync_single(struct device *hwdev, phys_addr_t tlb_addr,
swiotlb_full(struct device *dev, size_t size, enum dma_data_direction dir,
int do_panic)
{
+ if (swiotlb_force == SWIOTLB_NO_FORCE)
+ return;
+
/*
* Ran out of IOMMU space for this operation. This is very bad.
* Unfortunately the drivers cannot handle this operation properly.
--
1.9.1
At the end of the function, the local variable use_swiotlb has always
the same value as the global variable swiotlb. Hence drop the local
variable completely.
Signed-off-by: Geert Uytterhoeven <[email protected]>
---
v2:
- New.
---
arch/x86/kernel/pci-swiotlb.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/arch/x86/kernel/pci-swiotlb.c b/arch/x86/kernel/pci-swiotlb.c
index b47edb8f52566e22..36049af2715ebd13 100644
--- a/arch/x86/kernel/pci-swiotlb.c
+++ b/arch/x86/kernel/pci-swiotlb.c
@@ -68,12 +68,10 @@ void x86_swiotlb_free_coherent(struct device *dev, size_t size,
*/
int __init pci_swiotlb_detect_override(void)
{
- int use_swiotlb = swiotlb | swiotlb_force;
-
if (swiotlb_force)
swiotlb = 1;
- return use_swiotlb;
+ return swiotlb;
}
IOMMU_INIT_FINISH(pci_swiotlb_detect_override,
pci_xen_swiotlb_detect,
--
1.9.1
Convert the flag swiotlb_force from an int to an enum, to prepare for
the advent of more possible values.
Suggested-by: Konrad Rzeszutek Wilk <[email protected]>
Signed-off-by: Geert Uytterhoeven <[email protected]>
---
v2:
- New.
---
arch/arm64/mm/dma-mapping.c | 3 ++-
arch/arm64/mm/init.c | 3 ++-
arch/x86/kernel/pci-swiotlb.c | 2 +-
arch/x86/xen/pci-swiotlb-xen.c | 2 +-
drivers/xen/swiotlb-xen.c | 4 ++--
include/linux/swiotlb.h | 7 ++++++-
include/trace/events/swiotlb.h | 16 +++++++++-------
lib/swiotlb.c | 8 ++++----
8 files changed, 27 insertions(+), 18 deletions(-)
diff --git a/arch/arm64/mm/dma-mapping.c b/arch/arm64/mm/dma-mapping.c
index 290a84f3351f706b..e04082700bb16c35 100644
--- a/arch/arm64/mm/dma-mapping.c
+++ b/arch/arm64/mm/dma-mapping.c
@@ -524,7 +524,8 @@ struct dma_map_ops dummy_dma_ops = {
static int __init arm64_dma_init(void)
{
- if (swiotlb_force || max_pfn > (arm64_dma_phys_limit >> PAGE_SHIFT))
+ if (swiotlb_force == SWIOTLB_FORCE ||
+ max_pfn > (arm64_dma_phys_limit >> PAGE_SHIFT))
swiotlb = 1;
return atomic_pool_init();
diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c
index 212c4d1e2f26df7f..716d1226ba6925ba 100644
--- a/arch/arm64/mm/init.c
+++ b/arch/arm64/mm/init.c
@@ -401,7 +401,8 @@ static void __init free_unused_memmap(void)
*/
void __init mem_init(void)
{
- if (swiotlb_force || max_pfn > (arm64_dma_phys_limit >> PAGE_SHIFT))
+ if (swiotlb_force == SWIOTLB_FORCE ||
+ max_pfn > (arm64_dma_phys_limit >> PAGE_SHIFT))
swiotlb_init(1);
set_max_mapnr(pfn_to_page(max_pfn) - mem_map);
diff --git a/arch/x86/kernel/pci-swiotlb.c b/arch/x86/kernel/pci-swiotlb.c
index 36049af2715ebd13..410efb2c7b80028b 100644
--- a/arch/x86/kernel/pci-swiotlb.c
+++ b/arch/x86/kernel/pci-swiotlb.c
@@ -68,7 +68,7 @@ void x86_swiotlb_free_coherent(struct device *dev, size_t size,
*/
int __init pci_swiotlb_detect_override(void)
{
- if (swiotlb_force)
+ if (swiotlb_force == SWIOTLB_FORCE)
swiotlb = 1;
return swiotlb;
diff --git a/arch/x86/xen/pci-swiotlb-xen.c b/arch/x86/xen/pci-swiotlb-xen.c
index a9fafb5c873897ee..a0b36a9d5df149e6 100644
--- a/arch/x86/xen/pci-swiotlb-xen.c
+++ b/arch/x86/xen/pci-swiotlb-xen.c
@@ -48,7 +48,7 @@ int __init pci_xen_swiotlb_detect(void)
* activate this IOMMU. If running as PV privileged, activate it
* irregardless.
*/
- if ((xen_initial_domain() || swiotlb || swiotlb_force))
+ if (xen_initial_domain() || swiotlb || swiotlb_force == SWIOTLB_FORCE)
xen_swiotlb = 1;
/* If we are running under Xen, we MUST disable the native SWIOTLB.
diff --git a/drivers/xen/swiotlb-xen.c b/drivers/xen/swiotlb-xen.c
index 478fb91e3df2b8a9..aba12009422e343d 100644
--- a/drivers/xen/swiotlb-xen.c
+++ b/drivers/xen/swiotlb-xen.c
@@ -392,7 +392,7 @@ dma_addr_t xen_swiotlb_map_page(struct device *dev, struct page *page,
if (dma_capable(dev, dev_addr, size) &&
!range_straddles_page_boundary(phys, size) &&
!xen_arch_need_swiotlb(dev, phys, dev_addr) &&
- !swiotlb_force) {
+ (swiotlb_force != SWIOTLB_FORCE)) {
/* we are not interested in the dma_addr returned by
* xen_dma_map_page, only in the potential cache flushes executed
* by the function. */
@@ -552,7 +552,7 @@ void xen_swiotlb_unmap_page(struct device *hwdev, dma_addr_t dev_addr,
phys_addr_t paddr = sg_phys(sg);
dma_addr_t dev_addr = xen_phys_to_bus(paddr);
- if (swiotlb_force ||
+ if (swiotlb_force == SWIOTLB_FORCE ||
xen_arch_need_swiotlb(hwdev, paddr, dev_addr) ||
!dma_capable(hwdev, dev_addr, sg->length) ||
range_straddles_page_boundary(paddr, sg->length)) {
diff --git a/include/linux/swiotlb.h b/include/linux/swiotlb.h
index 183f37c8a5e164ad..71d104e4c8490763 100644
--- a/include/linux/swiotlb.h
+++ b/include/linux/swiotlb.h
@@ -9,7 +9,12 @@
struct page;
struct scatterlist;
-extern int swiotlb_force;
+enum swiotlb_force {
+ SWIOTLB_NORMAL, /* Default - depending on HW DMA mask etc. */
+ SWIOTLB_FORCE, /* swiotlb=force */
+};
+
+extern enum swiotlb_force swiotlb_force;
/*
* Maximum allowable number of contiguous slabs to map,
diff --git a/include/trace/events/swiotlb.h b/include/trace/events/swiotlb.h
index 7ea4c5e7c4487a96..5e2e30a7efce0ebd 100644
--- a/include/trace/events/swiotlb.h
+++ b/include/trace/events/swiotlb.h
@@ -11,16 +11,16 @@
TP_PROTO(struct device *dev,
dma_addr_t dev_addr,
size_t size,
- int swiotlb_force),
+ enum swiotlb_force swiotlb_force),
TP_ARGS(dev, dev_addr, size, swiotlb_force),
TP_STRUCT__entry(
- __string( dev_name, dev_name(dev) )
- __field( u64, dma_mask )
- __field( dma_addr_t, dev_addr )
- __field( size_t, size )
- __field( int, swiotlb_force )
+ __string( dev_name, dev_name(dev) )
+ __field( u64, dma_mask )
+ __field( dma_addr_t, dev_addr )
+ __field( size_t, size )
+ __field( enum swiotlb_force, swiotlb_force )
),
TP_fast_assign(
@@ -37,7 +37,9 @@
__entry->dma_mask,
(unsigned long long)__entry->dev_addr,
__entry->size,
- __entry->swiotlb_force ? "swiotlb_force" : "" )
+ __print_symbolic(__entry->swiotlb_force,
+ { SWIOTLB_NORMAL, "NORMAL" },
+ { SWIOTLB_FORCE, "FORCE" }))
);
#endif /* _TRACE_SWIOTLB_H */
diff --git a/lib/swiotlb.c b/lib/swiotlb.c
index cb1b54ee8527241d..a32dce6d5101f408 100644
--- a/lib/swiotlb.c
+++ b/lib/swiotlb.c
@@ -53,7 +53,7 @@
*/
#define IO_TLB_MIN_SLABS ((1<<20) >> IO_TLB_SHIFT)
-int swiotlb_force;
+enum swiotlb_force swiotlb_force;
/*
* Used to do a quick range check in swiotlb_tbl_unmap_single and
@@ -107,7 +107,7 @@
if (*str == ',')
++str;
if (!strcmp(str, "force"))
- swiotlb_force = 1;
+ swiotlb_force = SWIOTLB_FORCE;
return 0;
}
@@ -763,7 +763,7 @@ dma_addr_t swiotlb_map_page(struct device *dev, struct page *page,
* we can safely return the device addr and not worry about bounce
* buffering it.
*/
- if (dma_capable(dev, dev_addr, size) && !swiotlb_force)
+ if (dma_capable(dev, dev_addr, size) && swiotlb_force != SWIOTLB_FORCE)
return dev_addr;
trace_swiotlb_bounced(dev, dev_addr, size, swiotlb_force);
@@ -904,7 +904,7 @@ void swiotlb_unmap_page(struct device *hwdev, dma_addr_t dev_addr,
phys_addr_t paddr = sg_phys(sg);
dma_addr_t dev_addr = phys_to_dma(hwdev, paddr);
- if (swiotlb_force ||
+ if (swiotlb_force == SWIOTLB_FORCE ||
!dma_capable(hwdev, dev_addr, sg->length)) {
phys_addr_t map = map_single(hwdev, sg_phys(sg),
sg->length, dir, attrs);
--
1.9.1
On Fri, Dec 16, 2016 at 02:28:39PM +0100, Geert Uytterhoeven wrote:
> Hi Konrad,
Heya!
>
> This patch series adds a kernel command line option to aid debugging
> when developing support for DMA to memory outside the 32-bit address
> space. If specified, trying to map memory that cannot be used with DMA
> will fail, and a warning will be printed. This has been used
> succesfully to find drivers and DMA engines that do not support 64-bit
> memory.
>
> Changes compared to v1:
> - Dropp patch "swiotlb: Rate-limit printing when running out of
> SW-IOMMU space", which has been picked up,
> - Add cleanup patch 1/3,
> - Add patch 2/3, to convert the flag swiotlb_force from an int to an
> enum,
> - Change the kernel parameter "swiotlb=nobounce" to "swiotlb=noforce"
> (requested by Konrad Wilk),
> - Extend swiotlb_force enum instead of adding an swiotlb_nobounce
> variable (requested by Konrad Wilk).
>
> This has been tested on ra7796/salvator-x, and compile-tested on x86.
Let me run it on various combinations of x86 and if all is good
will ask Linus to pick it up for rc1.
Thanks!
>
> Thanks!
>
> Geert Uytterhoeven (3):
> x86, swiotlb: Simplify pci_swiotlb_detect_override()
> swiotlb: Convert swiotlb_force from int to enum
> swiotlb: Add swiotlb=noforce debug option
>
> Documentation/admin-guide/kernel-parameters.txt | 3 ++-
> arch/arm64/mm/dma-mapping.c | 3 ++-
> arch/arm64/mm/init.c | 3 ++-
> arch/x86/kernel/pci-swiotlb.c | 6 ++----
> arch/x86/xen/pci-swiotlb-xen.c | 2 +-
> drivers/xen/swiotlb-xen.c | 4 ++--
> include/linux/swiotlb.h | 8 +++++++-
> include/trace/events/swiotlb.h | 17 +++++++++-------
> lib/swiotlb.c | 26 +++++++++++++++++++------
> 9 files changed, 48 insertions(+), 24 deletions(-)
>
> --
> 1.9.1
>
> Gr{oetje,eeting}s,
>
> Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
>
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
> -- Linus Torvalds
On Fri, Dec 16, 2016 at 02:28:42PM +0100, Geert Uytterhoeven wrote:
> On architectures like arm64, swiotlb is tied intimately to the core
> architecture DMA support. In addition, ZONE_DMA cannot be disabled.
>
> To aid debugging and catch devices not supporting DMA to memory outside
> the 32-bit address space, add a kernel command line option
> "swiotlb=noforce", which disables the use of bounce buffers.
> If specified, trying to map memory that cannot be used with DMA will
> fail, and a rate-limited warning will be printed.
>
> Note that io_tlb_nslabs is set to 1, which is the minimal supported
> value.
>
> Signed-off-by: Geert Uytterhoeven <[email protected]>
What is this based on? I can't apply it on my latest that I had
sent to Linus?
Could you rebase this one please on:
git://git.kernel.org/pub/scm/linux/kernel/git/konrad/swiotlb.git stable/for-linus-4.9
Thanks?
Or if alternatively that does not - then please rebase it on
b5cab0da75c292ffa0fbd68dd2c820066b2842de
Thanks!
Hi Konrad,
On Mon, Dec 19, 2016 at 2:38 PM, Konrad Rzeszutek Wilk
<[email protected]> wrote:
> On Fri, Dec 16, 2016 at 02:28:42PM +0100, Geert Uytterhoeven wrote:
>> On architectures like arm64, swiotlb is tied intimately to the core
>> architecture DMA support. In addition, ZONE_DMA cannot be disabled.
>>
>> To aid debugging and catch devices not supporting DMA to memory outside
>> the 32-bit address space, add a kernel command line option
>> "swiotlb=noforce", which disables the use of bounce buffers.
>> If specified, trying to map memory that cannot be used with DMA will
>> fail, and a rate-limited warning will be printed.
>>
>> Note that io_tlb_nslabs is set to 1, which is the minimal supported
>> value.
>>
>> Signed-off-by: Geert Uytterhoeven <[email protected]>
>
> What is this based on? I can't apply it on my latest that I had
> sent to Linus?
I rebased it on last Friday's linux-next, due to the recent move of
kernel-parameters.txt.
All the rest should be identical to your linux-next branch.
> Could you rebase this one please on:
>
> git://git.kernel.org/pub/scm/linux/kernel/git/konrad/swiotlb.git stable/for-linus-4.9
That one indeed doesn't have the move of kernel-parameters.txt.
> Or if alternatively that does not - then please rebase it on
> b5cab0da75c292ffa0fbd68dd2c820066b2842de
However, my series already applies cleanly with git am on top of
that commit?
Please let me know what to do. Thanks!
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
On Mon, Dec 19, 2016 at 02:57:00PM +0100, Geert Uytterhoeven wrote:
> Hi Konrad,
>
> On Mon, Dec 19, 2016 at 2:38 PM, Konrad Rzeszutek Wilk
> <[email protected]> wrote:
> > On Fri, Dec 16, 2016 at 02:28:42PM +0100, Geert Uytterhoeven wrote:
> >> On architectures like arm64, swiotlb is tied intimately to the core
> >> architecture DMA support. In addition, ZONE_DMA cannot be disabled.
> >>
> >> To aid debugging and catch devices not supporting DMA to memory outside
> >> the 32-bit address space, add a kernel command line option
> >> "swiotlb=noforce", which disables the use of bounce buffers.
> >> If specified, trying to map memory that cannot be used with DMA will
> >> fail, and a rate-limited warning will be printed.
> >>
> >> Note that io_tlb_nslabs is set to 1, which is the minimal supported
> >> value.
> >>
> >> Signed-off-by: Geert Uytterhoeven <[email protected]>
> >
> > What is this based on? I can't apply it on my latest that I had
> > sent to Linus?
>
> I rebased it on last Friday's linux-next, due to the recent move of
> kernel-parameters.txt.
> All the rest should be identical to your linux-next branch.
>
> > Could you rebase this one please on:
> >
> > git://git.kernel.org/pub/scm/linux/kernel/git/konrad/swiotlb.git stable/for-linus-4.9
>
> That one indeed doesn't have the move of kernel-parameters.txt.
>
> > Or if alternatively that does not - then please rebase it on
> > b5cab0da75c292ffa0fbd68dd2c820066b2842de
>
> However, my series already applies cleanly with git am on top of
> that commit?
Ah that does indeed work.
Which means I can't push it to Linus until rc2 at least.
Linus is sad when he sees git pulls that hadn't "soaked"
for two weeks in linux-next - and especially patches on top of
merges during the merge window.
>
> Please let me know what to do. Thanks!
Building and testing them now.
Wish there was some travis script to do this.