2012-10-29 08:43:12

by Wen Congyang

[permalink] [raw]
Subject: [PATCH v3 0/2] fixes for mem= option

The documentation and implementation of 'mem=' option doesn't match, and the
option can't work for efi platform. This patchset updates the documentation
and make the option to work for efi platform.

Changes from v2 to v3:
Patch1: X86-32 ===> X86

Changes from v1 to v2:
Patch1: Just fix a typo error(ingoring -> ignoring).

Wen Congyang (2):
update mem= option's spec according to its implementation
x86: make 'mem=' option to work for efi platform

Documentation/kernel-parameters.txt | 7 ++++---
arch/x86/kernel/e820.c | 29 +++++++++++++++++++++++++----
2 files changed, 29 insertions(+), 7 deletions(-)

--
1.8.0


2012-10-29 08:43:09

by Wen Congyang

[permalink] [raw]
Subject: [PATCH v3 1/2] update mem= option's spec according to its implementation

Current mem= implementation seems buggy because specification and
implementation doesn't match. Current mem= has been working
for many years and it's not buggy, it works as expected. So
we should update the specification.

Signed-off-by: Wen Congyang <[email protected]>
Sort-of-tentatively-acked-by: Rob Landley <[email protected]>
---
Documentation/kernel-parameters.txt | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index 9776f06..91863a5 100644
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -1481,9 +1481,10 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
mem=nn[KMG] [KNL,BOOT] Force usage of a specific amount of memory
Amount of memory to be used when the kernel is not able
to see the whole system memory or for test.
- [X86-32] Use together with memmap= to avoid physical
- address space collisions. Without memmap= PCI devices
- could be placed at addresses belonging to unused RAM.
+ [X86] Work as limiting max address. Use together
+ with memmap= to avoid physical address space collisions.
+ Without memmap= PCI devices could be placed at addresses
+ belonging to unused RAM.

mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel
memory.
--
1.8.0

2012-10-29 08:43:34

by Wen Congyang

[permalink] [raw]
Subject: [PATCH v3 2/2] x86: make 'mem=' option to work for efi platform

Current mem boot option only can work for non efi environment. If the user
specifies add_efi_memmap, it cannot work for efi environment. In
the efi environment, we call e820_add_region() to add the memory map. So
we can modify __e820_add_region() and the mem boot option can work for
efi environment.

Signed-off-by: Wen Congyang <[email protected]>
---
arch/x86/kernel/e820.c | 29 +++++++++++++++++++++++++----
1 file changed, 25 insertions(+), 4 deletions(-)

diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c
index df06ade..b15b341 100644
--- a/arch/x86/kernel/e820.c
+++ b/arch/x86/kernel/e820.c
@@ -47,6 +47,7 @@ unsigned long pci_mem_start = 0xaeedbabe;
#ifdef CONFIG_PCI
EXPORT_SYMBOL(pci_mem_start);
#endif
+static u64 mem_limit = ~0ULL;

/*
* This function checks if any part of the range <start,end> is mapped
@@ -119,6 +120,20 @@ static void __init __e820_add_region(struct e820map *e820x, u64 start, u64 size,
return;
}

+ if (start >= mem_limit) {
+ printk(KERN_ERR "e820: ignoring [mem %#010llx-%#010llx]\n",
+ (unsigned long long)start,
+ (unsigned long long)(start + size - 1));
+ return;
+ }
+
+ if (mem_limit - start < size) {
+ printk(KERN_ERR "e820: ignoring [mem %#010llx-%#010llx]\n",
+ (unsigned long long)mem_limit,
+ (unsigned long long)(start + size - 1));
+ size = mem_limit - start;
+ }
+
e820x->map[x].addr = start;
e820x->map[x].size = size;
e820x->map[x].type = type;
@@ -809,7 +824,7 @@ static int userdef __initdata;
/* "mem=nopentium" disables the 4MB page tables. */
static int __init parse_memopt(char *p)
{
- u64 mem_size;
+ char *oldp;

if (!p)
return -EINVAL;
@@ -825,11 +840,11 @@ static int __init parse_memopt(char *p)
}

userdef = 1;
- mem_size = memparse(p, &p);
+ oldp = p;
+ mem_limit = memparse(p, &p);
/* don't remove all of memory when handling "mem={invalid}" param */
- if (mem_size == 0)
+ if (mem_limit == 0 || p == oldp)
return -EINVAL;
- e820_remove_range(mem_size, ULLONG_MAX - mem_size, E820_RAM, 1);

return 0;
}
@@ -881,6 +896,12 @@ early_param("memmap", parse_memmap_opt);

void __init finish_e820_parsing(void)
{
+ if (mem_limit != ~0ULL) {
+ userdef = 1;
+ e820_remove_range(mem_limit, ULLONG_MAX - mem_limit,
+ E820_RAM, 1);
+ }
+
if (userdef) {
u32 nr = e820.nr_map;

--
1.8.0

2012-10-29 10:48:12

by Richard Weinberger

[permalink] [raw]
Subject: Re: [PATCH v3 1/2] update mem= option's spec according to its implementation

On Mon, Oct 29, 2012 at 9:48 AM, Wen Congyang <[email protected]> wrote:
> Current mem= implementation seems buggy because specification and
> implementation doesn't match. Current mem= has been working
> for many years and it's not buggy, it works as expected. So
> we should update the specification.
>
> Signed-off-by: Wen Congyang <[email protected]>
> Sort-of-tentatively-acked-by: Rob Landley <[email protected]>

So, is this an ACK or not?

--
Thanks,
//richard

2012-10-29 10:54:58

by Wen Congyang

[permalink] [raw]
Subject: Re: [PATCH v3 1/2] update mem= option's spec according to its implementation

At 10/29/2012 06:48 PM, richard -rw- weinberger Wrote:
> On Mon, Oct 29, 2012 at 9:48 AM, Wen Congyang <[email protected]> wrote:
>> Current mem= implementation seems buggy because specification and
>> implementation doesn't match. Current mem= has been working
>> for many years and it's not buggy, it works as expected. So
>> we should update the specification.
>>
>> Signed-off-by: Wen Congyang <[email protected]>
>> Sort-of-tentatively-acked-by: Rob Landley <[email protected]>
>
> So, is this an ACK or not?
>

I don't know.

Here is the origin message:

At 06/15/2012 04:22 AM, Rob Landley Wrote:
> I have no objection to this but can't confirm it's true or not without
> an awful lot more digging through the code I don't have time for right
> now. (All the x86-32 machines I've used just had the 640k->1m hole and
> the rest was contiguous memory, so the behavior would be the same either
> way...)
>
> Sort-of-tentatively-acked-by: Rob Landley <[email protected]>
>
> Rob