Resend the v3 patch after reviewed by KAMEZAWA Hiroyuki. We still keep the
Old e820map, update variable max_pfn, max_low_pfn and high_memory only.
It is dependent on Fenguang's page_is_ram patch.
Memory-Hotplug: Fix the bug on interface /dev/mem for 64-bit kernel
The new added memory can not be access by interface /dev/mem, because we do not
update the variable high_memory, max_pfn and max_low_pfn.
Memory hotplug still has critical issues for 32-bit kernel, and it is more
important for 64-bit kernel, we fix it on 64-bit first. We add a function
update_end_of_memory_vars in file arch/x86/mm/init.c to update these variables.
CC: Andi Kleen <[email protected]>
CC: Li Haicheng <[email protected]>
Reviewed-by: Wu Fengguang <[email protected]>
Reviewed-by: KAMEZAWA Hiroyuki <[email protected]>
Signed-off-by: Shaohui Zheng <[email protected]>
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index d406c52..b6a85cc 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -1,6 +1,7 @@
#include <linux/initrd.h>
#include <linux/ioport.h>
#include <linux/swap.h>
+#include <linux/bootmem.h>
#include <asm/cacheflush.h>
#include <asm/e820.h>
@@ -386,3 +387,24 @@ void free_initrd_mem(unsigned long start, unsigned long end)
free_init_pages("initrd memory", start, end);
}
#endif
+
+/**
+ * After memory hotplug, the variable max_pfn, max_low_pfn and high_memory will
+ * be affected, it will be updated in this function. Memory hotplug still has
+ * critical issues on 32-bit kennel, it was more important on 64-bit kernel,
+ * so we update the variables for 64-bit kernel first, fix me in future for
+ * 32-bit kenrel.
+ */
+void __meminit __attribute__((weak)) update_end_of_memory_vars(u64 start,
+ u64 size)
+{
+#ifdef CONFIG_X86_64
+ unsigned long start_pfn = start >> PAGE_SHIFT;
+ unsigned long end_pfn = PFN_UP(start + size);
+
+ if (end_pfn > max_pfn) {
+ max_low_pfn = max_pfn = end_pfn;
+ high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
+ }
+#endif /* CONFIG_X86_64 */
+}
diff --git a/include/linux/bootmem.h b/include/linux/bootmem.h
index b10ec49..84533a5 100644
--- a/include/linux/bootmem.h
+++ b/include/linux/bootmem.h
@@ -13,6 +13,7 @@
extern unsigned long max_low_pfn;
extern unsigned long min_low_pfn;
+extern void update_end_of_memory_vars(u64 start, u64 size);
/*
* highest page
diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
index 030ce8a..3e94b23 100644
--- a/mm/memory_hotplug.c
+++ b/mm/memory_hotplug.c
@@ -523,6 +523,9 @@ int __ref add_memory(int nid, u64 start, u64 size)
BUG_ON(ret);
}
+ /* update max_pfn, max_low_pfn and high_memory */
+ update_end_of_memory_vars(start, size);
+
goto out;
error:
Thanks & Regards,
Shaohui
On Tue, 12 Jan 2010 15:45:54 +0800
"Zheng, Shaohui" <[email protected]> wrote:
> Resend the v3 patch after reviewed by KAMEZAWA Hiroyuki. We still keep the
> Old e820map, update variable max_pfn, max_low_pfn and high_memory only.
> It is dependent on Fenguang's page_is_ram patch.
>
> Memory-Hotplug: Fix the bug on interface /dev/mem for 64-bit kernel
>
> The new added memory can not be access by interface /dev/mem, because we do not
> update the variable high_memory, max_pfn and max_low_pfn.
>
> Memory hotplug still has critical issues for 32-bit kernel, and it is more
> important for 64-bit kernel, we fix it on 64-bit first. We add a function
> update_end_of_memory_vars in file arch/x86/mm/init.c to update these variables.
>
> CC: Andi Kleen <[email protected]>
> CC: Li Haicheng <[email protected]>
> Reviewed-by: Wu Fengguang <[email protected]>
> Reviewed-by: KAMEZAWA Hiroyuki <[email protected]>
> Signed-off-by: Shaohui Zheng <[email protected]>
3 points...
1. I think this patch cannot be compiled in archs other than x86. Right ?
IOW, please add static inline dummy...
2. pgdat->[start,end], totalram_pages etc...are updated at memory hotplug.
Please place the hook nearby them.
3. I recommend you yo use memory hotplug notifier.
If it's allowed, it will be cleaner.
It seems there are no strict ordering to update parameters this patch touches.
Thanks,
-Kame
> diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
> index d406c52..b6a85cc 100644
> --- a/arch/x86/mm/init.c
> +++ b/arch/x86/mm/init.c
> @@ -1,6 +1,7 @@
> #include <linux/initrd.h>
> #include <linux/ioport.h>
> #include <linux/swap.h>
> +#include <linux/bootmem.h>
>
> #include <asm/cacheflush.h>
> #include <asm/e820.h>
> @@ -386,3 +387,24 @@ void free_initrd_mem(unsigned long start, unsigned long end)
> free_init_pages("initrd memory", start, end);
> }
> #endif
> +
> +/**
> + * After memory hotplug, the variable max_pfn, max_low_pfn and high_memory will
> + * be affected, it will be updated in this function. Memory hotplug still has
> + * critical issues on 32-bit kennel, it was more important on 64-bit kernel,
> + * so we update the variables for 64-bit kernel first, fix me in future for
> + * 32-bit kenrel.
> + */
> +void __meminit __attribute__((weak)) update_end_of_memory_vars(u64 start,
> + u64 size)
> +{
> +#ifdef CONFIG_X86_64
> + unsigned long start_pfn = start >> PAGE_SHIFT;
> + unsigned long end_pfn = PFN_UP(start + size);
> +
> + if (end_pfn > max_pfn) {
> + max_low_pfn = max_pfn = end_pfn;
> + high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
> + }
> +#endif /* CONFIG_X86_64 */
> +}
> diff --git a/include/linux/bootmem.h b/include/linux/bootmem.h
> index b10ec49..84533a5 100644
> --- a/include/linux/bootmem.h
> +++ b/include/linux/bootmem.h
> @@ -13,6 +13,7 @@
>
> extern unsigned long max_low_pfn;
> extern unsigned long min_low_pfn;
> +extern void update_end_of_memory_vars(u64 start, u64 size);
>
> /*
> * highest page
> diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
> index 030ce8a..3e94b23 100644
> --- a/mm/memory_hotplug.c
> +++ b/mm/memory_hotplug.c
> @@ -523,6 +523,9 @@ int __ref add_memory(int nid, u64 start, u64 size)
> BUG_ON(ret);
> }
>
> + /* update max_pfn, max_low_pfn and high_memory */
> + update_end_of_memory_vars(start, size);
> +
> goto out;
>
> error:
>
> Thanks & Regards,
> Shaohui
>
>
>
3 points...
1. I think this patch cannot be compiled in archs other than x86. Right ?
IOW, please add static inline dummy...
[Zheng, Shaohui] Agree, I will add a static dummy function
2. pgdat->[start,end], totalram_pages etc...are updated at memory hotplug.
Please place the hook nearby them.
[Zheng, Shaohui] Agree.
3. I recommend you yo use memory hotplug notifier.
If it's allowed, it will be cleaner.
It seems there are no strict ordering to update parameters this patch touches.
[Zheng, Shaohui] Kame, do you means put the hook into function slab_mem_going_online_callback, it seems a good idea. If we select this method, we will need not to update these variable in function add_memory explicitly.
Thanks,
-Kame
On Tue, 12 Jan 2010 16:57:40 +0800
"Zheng, Shaohui" <[email protected]> wrote:
>
> 3 points...
> 1. I think this patch cannot be compiled in archs other than x86. Right ?
> IOW, please add static inline dummy...
> [Zheng, Shaohui] Agree, I will add a static dummy function
>
> 2. pgdat->[start,end], totalram_pages etc...are updated at memory hotplug.
> Please place the hook nearby them.
> [Zheng, Shaohui] Agree.
>
> 3. I recommend you yo use memory hotplug notifier.
> If it's allowed, it will be cleaner.
> It seems there are no strict ordering to update parameters this patch touches.
>
> [Zheng, Shaohui] Kame, do you means put the hook into function slab_mem_going_online_callback, it seems a good idea. If we select this method, we will need not to update these variable in function add_memory explicitly.
>
yes. I think callback is the best.
Thanks,
-Kame
>
> 3 points...
> 1. I think this patch cannot be compiled in archs other than x86. Right ?
> IOW, please add static inline dummy...
> [Zheng, Shaohui] Agree, I will add a static dummy function
>
> 2. pgdat->[start,end], totalram_pages etc...are updated at memory hotplug.
> Please place the hook nearby them.
> [Zheng, Shaohui] Agree.
>
> 3. I recommend you yo use memory hotplug notifier.
> If it's allowed, it will be cleaner.
> It seems there are no strict ordering to update parameters this patch touches.
>
> [Zheng, Shaohui] Kame, do you means put the hook into function slab_mem_going_online_callback, it seems a good idea. If we select this method, we will need not to update these variable in function add_memory explicitly.
>
yes. I think callback is the best.
[Zheng, Shaohui] it is okay for me, I will rewrite my patch and test it in local, thanks Kame :).
Thanks,
-Kame
Thanks & Regards,
Shaohui
>
> 3 points...
> 1. I think this patch cannot be compiled in archs other than x86. Right ?
> IOW, please add static inline dummy...
> [Zheng, Shaohui] Agree, I will add a static dummy function
>
> 2. pgdat->[start,end], totalram_pages etc...are updated at memory hotplug.
> Please place the hook nearby them.
> [Zheng, Shaohui] Agree.
>
> 3. I recommend you yo use memory hotplug notifier.
> If it's allowed, it will be cleaner.
> It seems there are no strict ordering to update parameters this patch touches.
>
> [Zheng, Shaohui] Kame, do you means put the hook into function slab_mem_going_online_callback, it seems a good idea. If we select this method, we will need not to update these variable in function add_memory explicitly.
>
yes. I think callback is the best.
[Zheng, Shaohui] Kame, I check the code carefully and write a simple patch, I found the callback will never be executed. The only entrance of this callback is in function online_pages/offline_pages, but I did not found explicit call to these 2 functions. Do we need to add a memory_notify to trigger this callback in function add_memory?
Thanks,
-Kame
Thanks & Regards,
Shaohui
Kame,
On Tue, Jan 12, 2010 at 05:04:33PM +0900, KAMEZAWA Hiroyuki wrote:
> 3 points...
> 1. I think this patch cannot be compiled in archs other than x86. Right ?
> IOW, please add static inline dummy...
Good catch!
> 2. pgdat->[start,end], totalram_pages etc...are updated at memory hotplug.
> Please place the hook nearby them.
arch/x86/mm/init_64.c:arch_add_memory() updates max_pfn_mapped, in
this sense it's equally OK to update max_pfn/max_low_pfn etc before
the call to arch_add_memory() ;)
> 3. I recommend you yo use memory hotplug notifier.
> If it's allowed, it will be cleaner.
Hmm, notifier is for _outsider_ subsystems. It smells a bit
overkill to do notifier _inside_ the hotplug code.
> It seems there are no strict ordering to update parameters this patch touches.
I tend to agree. That said, it does help keep our mind straight if we do it
in some logical order: max_pfn => max_pfn_mapped => totalram_pages etc.
Thanks,
Fengguang
> > 2. pgdat->[start,end], totalram_pages etc...are updated at memory hotplug.
> > Please place the hook nearby them.
>
> arch/x86/mm/init_64.c:arch_add_memory() updates max_pfn_mapped, in
> this sense it's equally OK to update max_pfn/max_low_pfn etc before
> the call to arch_add_memory() ;)
Shaohui, I'd suggest to update max_pfn/max_low_pfn/high_memory in
arch/x86/mm/init_64.c:arch_add_memory() now, for X86_64.
Later on we can add code to arch/x86/mm/init_32.c:arch_add_memory()
for X86_32.
The code cannot be shared anyway.
Thanks,
Fengguang