ZONE_DMA is unnecessary for a large number of machines that do not
require addressing in the lower 16MB of memory because they do not use
ISA devices with 16-bit address registers (plus one page byte register).
This patch allows users to disable ZONE_DMA for x86 if they know they
will not be using such devices with their kernel.
This prevents the VM from unnecessarily reserving a ratio of memory
(defaulting to 1/256th of system capacity) with lowmem_reserve_ratio
for such allocations when it will never be used.
Signed-off-by: David Rientjes <[email protected]>
---
v2: updated from October 2010
H. Peter Anvin requested that CONFIG_EMBEDDED be renamed to
CONFIG_EXPERT prior to this change in the original thread:
http://marc.info/?t=128701541500005
I did that at Ingo's suggestion. Also, CONFIG_GENERIC_ISA_DMA
and CONFIG_ISA_DMA_API are now configurable on x86 as discussed
in the same thread, as well, for additional text savings in
combination with this patch.
arch/x86/Kconfig | 9 ++++++++-
arch/x86/mm/init_32.c | 2 ++
arch/x86/mm/init_64.c | 2 ++
3 files changed, 12 insertions(+), 1 deletions(-)
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -112,7 +112,14 @@ config MMU
def_bool y
config ZONE_DMA
- def_bool y
+ bool "DMA memory allocation support" if EXPERT
+ default y
+ help
+ DMA memory allocation support allows old ISA devices with 16-bit
+ address registers to allocate memory within the first 16MB of
+ address space. Disable if no such devices will be used.
+
+ If unsure, say Y.
config SBUS
bool
diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c
--- a/arch/x86/mm/init_32.c
+++ b/arch/x86/mm/init_32.c
@@ -678,8 +678,10 @@ static void __init zone_sizes_init(void)
{
unsigned long max_zone_pfns[MAX_NR_ZONES];
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+#ifdef CONFIG_ZONE_DMA
max_zone_pfns[ZONE_DMA] =
virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
+#endif
max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
#ifdef CONFIG_HIGHMEM
max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -616,7 +616,9 @@ void __init paging_init(void)
unsigned long max_zone_pfns[MAX_NR_ZONES];
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+#ifdef CONFIG_ZONE_DMA
max_zone_pfns[ZONE_DMA] = MAX_DMA_PFN;
+#endif
max_zone_pfns[ZONE_DMA32] = MAX_DMA32_PFN;
max_zone_pfns[ZONE_NORMAL] = max_pfn;
David Rientjes wrote:
> ZONE_DMA is unnecessary for a large number of machines that do not
> require addressing in the lower 16MB of memory because they do not use
> ISA devices with 16-bit address registers (plus one page byte register).
>
> This patch allows users to disable ZONE_DMA for x86 if they know they
> will not be using such devices with their kernel.
> ...
> config ZONE_DMA
> - def_bool y
> + bool "DMA memory allocation support" if EXPERT
Better call it "ISA/LPC DMA memory allocation support".
> + default y
Does it make sense to have this without CONFIG_ISA_DMA_API?
> + help
> + DMA memory allocation support allows old ISA devices with 16-bit
> + address registers to allocate memory within the first 16MB of
> + address space. Disable if no such devices will be used.
"Old ISA devices" might be misleading because ISA does not imply actual
ISA slots; please mention LPC devices like the floppy controller and the
parallel port in ECP mode.
Regards,
Clemens
On 04/15/2011 04:42 AM, Clemens Ladisch wrote:
> David Rientjes wrote:
>> ZONE_DMA is unnecessary for a large number of machines that do not
>> require addressing in the lower 16MB of memory because they do not use
>> ISA devices with 16-bit address registers (plus one page byte register).
>>
>> This patch allows users to disable ZONE_DMA for x86 if they know they
>> will not be using such devices with their kernel.
>> ...
>> config ZONE_DMA
>> - def_bool y
>> + bool "DMA memory allocation support" if EXPERT
>
> Better call it "ISA/LPC DMA memory allocation support".
>
>> + default y
>
> Does it make sense to have this without CONFIG_ISA_DMA_API?
>
>> + help
>> + DMA memory allocation support allows old ISA devices with 16-bit
>> + address registers to allocate memory within the first 16MB of
>> + address space. Disable if no such devices will be used.
>
> "Old ISA devices" might be misleading because ISA does not imply actual
> ISA slots; please mention LPC devices like the floppy controller and the
> parallel port in ECP mode.
>
This isn't really a very good edit, as ZONE_DMA isn't inherently limited
to ISA/LPC devices (and LPC devices which do LPC-native bus mastering
DMA are not affected.) ZONE_DMA is used even for some PCI devices which
have restrictions on the DMA address mask.
-hpa
On Fri, 15 Apr 2011, H. Peter Anvin wrote:
> On 04/15/2011 04:42 AM, Clemens Ladisch wrote:
> > David Rientjes wrote:
> >> ZONE_DMA is unnecessary for a large number of machines that do not
> >> require addressing in the lower 16MB of memory because they do not use
> >> ISA devices with 16-bit address registers (plus one page byte register).
> >>
> >> This patch allows users to disable ZONE_DMA for x86 if they know they
> >> will not be using such devices with their kernel.
> >> ...
> >> config ZONE_DMA
> >> - def_bool y
> >> + bool "DMA memory allocation support" if EXPERT
> >
> > Better call it "ISA/LPC DMA memory allocation support".
> >
> >> + default y
> >
> > Does it make sense to have this without CONFIG_ISA_DMA_API?
> >
> >> + help
> >> + DMA memory allocation support allows old ISA devices with 16-bit
> >> + address registers to allocate memory within the first 16MB of
> >> + address space. Disable if no such devices will be used.
> >
> > "Old ISA devices" might be misleading because ISA does not imply actual
> > ISA slots; please mention LPC devices like the floppy controller and the
> > parallel port in ECP mode.
> >
>
> This isn't really a very good edit, as ZONE_DMA isn't inherently limited
> to ISA/LPC devices (and LPC devices which do LPC-native bus mastering
> DMA are not affected.) ZONE_DMA is used even for some PCI devices which
> have restrictions on the DMA address mask.
>
Peter, do you think we can merge this patch now? The discussion from
October resulted in the decision to rename CONFIG_EMBEDDED to
CONFIG_EXPERT and make CONFIG_GENERIC_ISA_DMA and CONFIG_ISA_DMA_API
configurable for x86. That's all been done via the -mm tree and is in the
latest git.
We've run without ZONE_DMA for a few years without issues (and, obviously,
this patch) on x86.
On 04/20/2011 07:31 PM, David Rientjes wrote:
>>
>> This isn't really a very good edit, as ZONE_DMA isn't inherently limited
>> to ISA/LPC devices (and LPC devices which do LPC-native bus mastering
>> DMA are not affected.) ZONE_DMA is used even for some PCI devices which
>> have restrictions on the DMA address mask.
>>
>
> Peter, do you think we can merge this patch now? The discussion from
> October resulted in the decision to rename CONFIG_EMBEDDED to
> CONFIG_EXPERT and make CONFIG_GENERIC_ISA_DMA and CONFIG_ISA_DMA_API
> configurable for x86. That's all been done via the -mm tree and is in the
> latest git.
>
> We've run without ZONE_DMA for a few years without issues (and, obviously,
> this patch) on x86.
Please address the documentation issues. ZONE_DMA is *NOT* just for
"old ISA devices" -- there are plenty of other devices with restricted
DMA masks.
-hpa
--
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel. I don't speak on their behalf.
ZONE_DMA is unnecessary for a large number of machines that do not
require addressing in the lower 16MB of memory because they do not use
devices with 16-bit address registers (plus one page byte register).
This patch allows users to disable ZONE_DMA for x86 if they know they
will not be using such devices with their kernel.
This prevents the VM from unnecessarily reserving a ratio of memory
(defaulting to 1/256th of system capacity) with lowmem_reserve_ratio
for such allocations when it will never be used.
Signed-off-by: David Rientjes <[email protected]>
---
v3: did s/old ISA devices/devices/ as suggested by H. Peter Anvin.
arch/x86/Kconfig | 9 ++++++++-
arch/x86/mm/init_32.c | 2 ++
arch/x86/mm/init_64.c | 2 ++
3 files changed, 12 insertions(+), 1 deletions(-)
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -112,7 +112,14 @@ config MMU
def_bool y
config ZONE_DMA
- def_bool y
+ bool "DMA memory allocation support" if EXPERT
+ default y
+ help
+ DMA memory allocation support allows devices with 16-bit address
+ registers to allocate memory within the first 16MB of address space.
+ Disable if no such devices will be used.
+
+ If unsure, say Y.
config SBUS
bool
diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c
--- a/arch/x86/mm/init_32.c
+++ b/arch/x86/mm/init_32.c
@@ -678,8 +678,10 @@ static void __init zone_sizes_init(void)
{
unsigned long max_zone_pfns[MAX_NR_ZONES];
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+#ifdef CONFIG_ZONE_DMA
max_zone_pfns[ZONE_DMA] =
virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
+#endif
max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
#ifdef CONFIG_HIGHMEM
max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -616,7 +616,9 @@ void __init paging_init(void)
unsigned long max_zone_pfns[MAX_NR_ZONES];
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+#ifdef CONFIG_ZONE_DMA
max_zone_pfns[ZONE_DMA] = MAX_DMA_PFN;
+#endif
max_zone_pfns[ZONE_DMA32] = MAX_DMA32_PFN;
max_zone_pfns[ZONE_NORMAL] = max_pfn;
On Wed, 27 Apr 2011, David Rientjes wrote:
> ZONE_DMA is unnecessary for a large number of machines that do not
> require addressing in the lower 16MB of memory because they do not use
> devices with 16-bit address registers (plus one page byte register).
>
> This patch allows users to disable ZONE_DMA for x86 if they know they
> will not be using such devices with their kernel.
>
> This prevents the VM from unnecessarily reserving a ratio of memory
> (defaulting to 1/256th of system capacity) with lowmem_reserve_ratio
> for such allocations when it will never be used.
>
> Signed-off-by: David Rientjes <[email protected]>
Peter, is this ready for merge?
On 05/03/2011 05:00 PM, David Rientjes wrote:
> On Wed, 27 Apr 2011, David Rientjes wrote:
>
>> ZONE_DMA is unnecessary for a large number of machines that do not
>> require addressing in the lower 16MB of memory because they do not use
>> devices with 16-bit address registers (plus one page byte register).
>>
>> This patch allows users to disable ZONE_DMA for x86 if they know they
>> will not be using such devices with their kernel.
>>
>> This prevents the VM from unnecessarily reserving a ratio of memory
>> (defaulting to 1/256th of system capacity) with lowmem_reserve_ratio
>> for such allocations when it will never be used.
>>
>> Signed-off-by: David Rientjes <[email protected]>
>
> Peter, is this ready for merge?
+ DMA memory allocation support allows devices with 16-bit address
+ registers to allocate memory within the first 16MB of address space.
+ Disable if no such devices will be used.
This is still not correct... this is for devices with less than a 32-bit
mask, not just for things that have 24-bit (not 16-bit!) DMA address
restrictions.
Please get this right.
Other than that, I don't have any technical objections.
-hpa
ZONE_DMA is unnecessary for a large number of machines that do not
require addressing in the lower 16MB of memory because they do not use
devices with 16-bit address registers (plus one page byte register).
This patch allows users to disable ZONE_DMA for x86 if they know they
will not be using such devices with their kernel.
This prevents the VM from unnecessarily reserving a ratio of memory
(defaulting to 1/256th of system capacity) with lowmem_reserve_ratio
for such allocations when it will never be used.
Signed-off-by: David Rientjes <[email protected]>
---
arch/x86/Kconfig | 9 ++++++++-
arch/x86/mm/init_32.c | 2 ++
arch/x86/mm/init_64.c | 2 ++
3 files changed, 12 insertions(+), 1 deletions(-)
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -112,7 +112,14 @@ config MMU
def_bool y
config ZONE_DMA
- def_bool y
+ bool "DMA memory allocation support" if EXPERT
+ default y
+ help
+ DMA memory allocation support allows devices with less than 32-bit
+ addressing to allocate within the first 16MB of address space.
+ Disable if no such devices will be used.
+
+ If unsure, say Y.
config SBUS
bool
diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c
--- a/arch/x86/mm/init_32.c
+++ b/arch/x86/mm/init_32.c
@@ -678,8 +678,10 @@ static void __init zone_sizes_init(void)
{
unsigned long max_zone_pfns[MAX_NR_ZONES];
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+#ifdef CONFIG_ZONE_DMA
max_zone_pfns[ZONE_DMA] =
virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
+#endif
max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
#ifdef CONFIG_HIGHMEM
max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -616,7 +616,9 @@ void __init paging_init(void)
unsigned long max_zone_pfns[MAX_NR_ZONES];
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+#ifdef CONFIG_ZONE_DMA
max_zone_pfns[ZONE_DMA] = MAX_DMA_PFN;
+#endif
max_zone_pfns[ZONE_DMA32] = MAX_DMA32_PFN;
max_zone_pfns[ZONE_NORMAL] = max_pfn;
On 05/16/2011 01:54 PM, David Rientjes wrote:
> ZONE_DMA is unnecessary for a large number of machines that do not
> require addressing in the lower 16MB of memory because they do not use
> devices with 16-bit address registers (plus one page byte register).
You keep repeating this error, although this time you at least fixed the
Kconfig file. I'll fix up the checkin message and apply.
-hpa
Commit-ID: dc382fd5bcca7098a984705ed6ac880f539d068e
Gitweb: http://git.kernel.org/tip/dc382fd5bcca7098a984705ed6ac880f539d068e
Author: David Rientjes <[email protected]>
AuthorDate: Mon, 16 May 2011 13:54:10 -0700
Committer: H. Peter Anvin <[email protected]>
CommitDate: Mon, 16 May 2011 14:03:28 -0700
x86, mm: Allow ZONE_DMA to be configurable
ZONE_DMA is unnecessary for a large number of machines that do not
require less than 32-bit DMA addressing, e.g. ISA legacy DMA or PCI
cards with a restricted DMA address mask.
This patch allows users to disable ZONE_DMA for x86 if they know they
will not be using such devices with their kernel.
This prevents the VM from unnecessarily reserving a ratio of memory
(defaulting to 1/256th of system capacity) with lowmem_reserve_ratio
for such allocations when it will never be used.
Signed-off-by: David Rientjes <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: H. Peter Anvin <[email protected]>
---
arch/x86/Kconfig | 9 ++++++++-
arch/x86/mm/init_32.c | 2 ++
arch/x86/mm/init_64.c | 2 ++
3 files changed, 12 insertions(+), 1 deletions(-)
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 648fca4..0eb801a 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -112,7 +112,14 @@ config MMU
def_bool y
config ZONE_DMA
- def_bool y
+ bool "DMA memory allocation support" if EXPERT
+ default y
+ help
+ DMA memory allocation support allows devices with less than 32-bit
+ addressing to allocate within the first 16MB of address space.
+ Disable if no such devices will be used.
+
+ If unsure, say Y.
config SBUS
bool
diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c
index 2cde0a34..29f7c6d 100644
--- a/arch/x86/mm/init_32.c
+++ b/arch/x86/mm/init_32.c
@@ -678,8 +678,10 @@ static void __init zone_sizes_init(void)
{
unsigned long max_zone_pfns[MAX_NR_ZONES];
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+#ifdef CONFIG_ZONE_DMA
max_zone_pfns[ZONE_DMA] =
virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
+#endif
max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
#ifdef CONFIG_HIGHMEM
max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index 0404bb3..d865c4a 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -616,7 +616,9 @@ void __init paging_init(void)
unsigned long max_zone_pfns[MAX_NR_ZONES];
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+#ifdef CONFIG_ZONE_DMA
max_zone_pfns[ZONE_DMA] = MAX_DMA_PFN;
+#endif
max_zone_pfns[ZONE_DMA32] = MAX_DMA32_PFN;
max_zone_pfns[ZONE_NORMAL] = max_pfn;