mm/memblock: reuse memblock bottom-up allocation style
Subsystems (6)
v3 cover mm/memblock: reuse memblock bottom-up allocation style
The bottom-up allocation style is introduced to cope with movable_node, where the limit inferior of allocation starts from kernel's end, due to lack of knowledge of memory hotplug info at this early time. Beside this original aim, 'kexec -c' prefers to reuse this style to alloc mem at lower address, since if the reserved region is beyond 4G, then it requires extra mem (default is 16M) for swiotlb. But at this time hotplug info has been got, the limit inferior can be extend to 0, which is done by this series
Pingfan Liu (2):
mm/memblock: extend the limit inferior of bottom-up after parsing
hotplug attr
x86/kdump: bugfix, make the behavior of crashkernel=X consistent with
kaslr
arch/x86/kernel/setup.c | 9 +++++--- drivers/acpi/numa.c | 4 ++++ include/linux/memblock.h | 1 + mm/memblock.c | 58 +++++++++++++++++++++++++++++------------------- 4 files changed, 46 insertions(+), 26 deletions(-)
Signature · 1 line
2.7.4
PATCH v3 1/2 mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
The bottom-up allocation style is introduced to cope with movable_node, where the limit inferior of allocation starts from kernel's end, due to lack of knowledge of memory hotplug info at this early time. But if later, hotplug info has been got, the limit inferior can be extend to 0. 'kexec -c' prefers to reuse this style to alloc mem at lower address, since if the reserved region is beyond 4G, then it requires extra mem (default is 16M) for swiotlb.
--- drivers/acpi/numa.c | 4 ++++ include/linux/memblock.h | 1 + mm/memblock.c | 58 +++++++++++++++++++++++++++++------------------- 3 files changed, 40 insertions(+), 23 deletions(-)
drivers/acpi/numa.c +4 -0
@@ -462,6 +462,10 @@ int __init acpi_numa_init(void) cnt =acpi_table_parse_srat (ACPI_SRAT_TYPE_MEMORY_AFFINITY , acpi_parse_memory_affinity,0 ); + +#if defined(CONFIG_X86 ) || defined(CONFIG_ARM64 ) +mark_mem_hotplug_parsed (); +#endif }/* SLIT: System Locality Information Table */
include/linux/memblock.h +1 -0
@@ -125,6 +125,7 @@ int memblock_reserve(phys_addr_t base, phys_addr_t size);void memblock_trim_memory (phys_addr_t align);bool memblock_overlaps_region (struct memblock_type *type,phys_addr_t base,phys_addr_t size); +void mark_mem_hotplug_parsed (void );int memblock_mark_hotplug (phys_addr_t base,phys_addr_t size);int memblock_clear_hotplug (phys_addr_t base,phys_addr_t size);int memblock_mark_mirror (phys_addr_t base,phys_addr_t size);
mm/memblock.c +35 -23
82 lines not on this page · raw message
Signature · 1 line
2.7.4
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
Quoted 7 lines · Pingfan Liu
The bottom-up allocation style is introduced to cope with
movable_node, where the limit inferior of allocation starts from kernel's end, due to lack of knowledge of memory hotplug info at this early time. But if later, hotplug info has been got, the limit inferior can be extend to 0. 'kexec -c' prefers to reuse this style to alloc mem at lower address, since if the reserved region is beyond 4G, then it requires extra mem (default is 16M) for swiotlb.
I fail to understand why the availability of memory hotplug information would allow to extend the lower limit of bottom-up memblock allocations below the kernel. The memory in the physical range [0, kernel_start) can be allocated as soon as the kernel memory is reserved.
The extents of the memory node hosting the kernel image can be used to limit memblok allocations from that particular node, even in top-down mode.
Quoted 145 lines
Signed-off-by Pingfan Liu <[email protected]>Cc Tang Chen <[email protected]>Cc "Rafael J. Wysocki" <[email protected]>Cc Len Brown <[email protected]>Cc Andrew Morton <[email protected]>Cc Mike Rapoport <[email protected]>Cc Michal Hocko <[email protected]>Cc Jonathan Corbet <[email protected]>Cc Yaowei Bai <[email protected]>Cc Pavel Tatashin <[email protected]>Cc Nicholas Piggin <[email protected]>Cc Naoya Horiguchi <[email protected]>Cc Daniel Vacek <[email protected]>Cc Mathieu Malaterre <[email protected]>Cc Stefan Agner <[email protected]>Cc Dave Young <[email protected]>Cc Baoquan He <[email protected]>--- drivers/acpi/numa.c | 4 ++++ include/linux/memblock.h | 1 + mm/memblock.c | 58 +++++++++++++++++++++++++++++------------------- 3 files changed, 40 insertions(+), 23 deletions(-)drivers/acpi/numa.c +4
-0@@ -462,6 +462,10 @@ int __init acpi_numa_init(void) cnt =acpi_table_parse_srat (ACPI_SRAT_TYPE_MEMORY_AFFINITY , acpi_parse_memory_affinity,0 ); + +#if defined(CONFIG_X86 ) || defined(CONFIG_ARM64 ) +mark_mem_hotplug_parsed (); +#endif }/* SLIT: System Locality Information Table */ include/linux/memblock.h +1
-0@@ -125,6 +125,7 @@ int memblock_reserve(phys_addr_t base, phys_addr_t size);void memblock_trim_memory (phys_addr_t align);bool memblock_overlaps_region (struct memblock_type *type,phys_addr_t base,phys_addr_t size); +void mark_mem_hotplug_parsed (void );int memblock_mark_hotplug (phys_addr_t base,phys_addr_t size);int memblock_clear_hotplug (phys_addr_t base,phys_addr_t size);int memblock_mark_mirror (phys_addr_t base,phys_addr_t size);mm/memblock.c +35
-23@@ -231,6 +231,12 @@ __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,return 0 ; } +static bool mem_hotmovable_parsed __initdata_memblock; +void __init_memblock mark_mem_hotplug_parsed(void) +{ + mem_hotmovable_parsed = true; +} +/** * memblock_find_in_range_node - find free area in given range and node * @size: size of free area to find @@ -259,7 +265,7 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, phys_addr_t end, int nid, enum memblock_flags flags) { - phys_addr_t kernel_end, ret; + phys_addr_t kernel_end, ret = 0;/* pump up @end */ if (end ==MEMBLOCK_ALLOC_ACCESSIBLE )@@ -270,34 +276,40 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, end =max (start, end); kernel_end =__pa_symbol (_end); -/* -* try bottom-up allocation only when bottom-up mode -* is set and @end is above the kernel image. -*/ -if (memblock_bottom_up () && end > kernel_end) { -phys_addr_t bottom_up_start; +if (memblock_bottom_up ()) { +phys_addr_t bottom_up_start = start; -/* make sure we will allocate above the kernel */ - bottom_up_start =max (start, kernel_end); - -/* ok, try bottom-up allocation first */ - ret =__memblock_find_range_bottom_up (bottom_up_start, end, - size, align, nid, flags); -if (ret) +if (mem_hotmovable_parsed) { + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags);return ret;/* -* we always limit bottom-up allocation above the kernel, -* but top-down allocation doesn't have the limit, so -* retrying top-down allocation may succeed when bottom-up -* allocation failed. -* -* bottom-up allocation is expected to be fail very rarely, -* so we use WARN_ONCE() here to see the stack trace if -* fail happens. +* if mem hotplug info is not parsed yet, try bottom-up +* allocation with @end above the kernel image. */ -WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ), + }else if (!mem_hotmovable_parsed && end > kernel_end) { +/* make sure we will allocate above the kernel */ + bottom_up_start =max (start, kernel_end); + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags); +if (ret) +return ret; +/* +* we always limit bottom-up allocation above the +* kernel, but top-down allocation doesn't have +* the limit, so retrying top-down allocation may +* succeed when bottom-up allocation failed. +* +* bottom-up allocation is expected to be fail +* very rarely, so we use WARN_ONCE() here to see +* the stack trace if fail happens. +*/ +WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ),"memblock: bottom-up allocation failed, memory hotremove may be affected\n" ); + } } return __memblock_find_range_top_down(start, end, size, align, nid,Signature · 1 line
2.7.4
Signature · 2 lines
Sincerely yours, Mike.
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
Quoted 10 lines · Mike Rapoport
Quoted 7 lines · Pingfan Liu
The bottom-up allocation style is introduced to cope with
movable_node, where the limit inferior of allocation starts from kernel's end, due to lack of knowledge of memory hotplug info at this early time. But if later, hotplug info has been got, the limit inferior can be extend to 0. 'kexec -c' prefers to reuse this style to alloc mem at lower address, since if the reserved region is beyond 4G, then it requires extra mem (default is 16M) for swiotlb.
Quoted 5 lines
I fail to understand why the availability of memory hotplug information would allow to extend the lower limit of bottom-up memblock allocations below the kernel. The memory in the physical range [0,
kernel_start) can be allocated as soon as the kernel memory is reserved.
Yes, the [0, kernel_start) can be allocated at this time by some func e.g. memblock_reserve(). But there is trick. For the func like memblock_find_in_range(), this is hotplug attr checking ,,it will check the hotmovable attr in __next_mem_range()
{
if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) continue
}. So the movable memory can be safely skipped.
Thanks for your kindly review.
Regards,
Pingfan
Quoted 153 lines
The extents of the memory node hosting the kernel image can be used to limit memblok allocations from that particular node, even in top-down mode.
Quoted 143 lines
Signed-off-by Pingfan Liu <[email protected]>Cc Tang Chen <[email protected]>Cc "Rafael J. Wysocki" <[email protected]>Cc Len Brown <[email protected]>Cc Andrew Morton <[email protected]>Cc Mike Rapoport <[email protected]>Cc Michal Hocko <[email protected]>Cc Jonathan Corbet <[email protected]>Cc Yaowei Bai <[email protected]>Cc Pavel Tatashin <[email protected]>Cc Nicholas Piggin <[email protected]>Cc Naoya Horiguchi <[email protected]>Cc Daniel Vacek <[email protected]>Cc Mathieu Malaterre <[email protected]>Cc Stefan Agner <[email protected]>Cc Dave Young <[email protected]>Cc Baoquan He <[email protected]>--- drivers/acpi/numa.c | 4 ++++ include/linux/memblock.h | 1 + mm/memblock.c | 58 +++++++++++++++++++++++++++++------------------- 3 files changed, 40 insertions(+), 23 deletions(-)drivers/acpi/numa.c +4
-0@@ -462,6 +462,10 @@ int __init acpi_numa_init(void) cnt =acpi_table_parse_srat (ACPI_SRAT_TYPE_MEMORY_AFFINITY , acpi_parse_memory_affinity,0 ); + +#if defined(CONFIG_X86 ) || defined(CONFIG_ARM64 ) +mark_mem_hotplug_parsed (); +#endif }/* SLIT: System Locality Information Table */ include/linux/memblock.h +1
-0@@ -125,6 +125,7 @@ int memblock_reserve(phys_addr_t base, phys_addr_t size);void memblock_trim_memory (phys_addr_t align);bool memblock_overlaps_region (struct memblock_type *type,phys_addr_t base,phys_addr_t size); +void mark_mem_hotplug_parsed (void );int memblock_mark_hotplug (phys_addr_t base,phys_addr_t size);int memblock_clear_hotplug (phys_addr_t base,phys_addr_t size);int memblock_mark_mirror (phys_addr_t base,phys_addr_t size);mm/memblock.c +35
-24@@ -231,6 +231,12 @@ __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,return 0 ; } +static bool mem_hotmovable_parsed __initdata_memblock; +void __init_memblock mark_mem_hotplug_parsed(void) +{ + mem_hotmovable_parsed = true; +} +/** * memblock_find_in_range_node - find free area in given range and node * @size: size of free area to find @@ -259,7 +265,7 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, phys_addr_t end, int nid, enum memblock_flags flags) { - phys_addr_t kernel_end, ret; + phys_addr_t kernel_end, ret = 0;/* pump up @end */ if (end ==MEMBLOCK_ALLOC_ACCESSIBLE )@@ -270,34 +276,40 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, end =max (start, end); kernel_end =__pa_symbol (_end); -/* -* try bottom-up allocation only when bottom-up mode -* is set and @end is above the kernel image. -*/ -if (memblock_bottom_up () && end > kernel_end) { -phys_addr_t bottom_up_start; +if (memblock_bottom_up ()) { +phys_addr_t bottom_up_start = start; -/* make sure we will allocate above the kernel */ - bottom_up_start =max (start, kernel_end); - -/* ok, try bottom-up allocation first */ - ret =__memblock_find_range_bottom_up (bottom_up_start, end, - size, align, nid, flags); -if (ret) +if (mem_hotmovable_parsed) { + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags);return ret;/* -* we always limit bottom-up allocation above the kernel, -* but top-down allocation doesn't have the limit, so -* retrying top-down allocation may succeed when bottom-up -* allocation failed. -* -* bottom-up allocation is expected to be fail very rarely, -* so we use WARN_ONCE() here to see the stack trace if -* fail happens. +* if mem hotplug info is not parsed yet, try bottom-up +* allocation with @end above the kernel image. */ -WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ), + }else if (!mem_hotmovable_parsed && end > kernel_end) { +/* make sure we will allocate above the kernel */ + bottom_up_start =max (start, kernel_end); + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags); +if (ret) +return ret; +/* +* we always limit bottom-up allocation above the +* kernel, but top-down allocation doesn't have +* the limit, so retrying top-down allocation may +* succeed when bottom-up allocation failed. +* +* bottom-up allocation is expected to be fail +* very rarely, so we use WARN_ONCE() here to see +* the stack trace if fail happens. +*/ +WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ),"memblock: bottom-up allocation failed, memory hotremove may be affected\n" ); + } } return __memblock_find_range_top_down(start, end, size, align, nid, --Quoted 2 lines
2.7.4
--
Sincerely yours,
Mike.
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
Quoted 16 lines · Pingfan Liu
Quoted 10 lines · Mike Rapoport
Quoted 7 lines · Pingfan Liu
The bottom-up allocation style is introduced to cope with
movable_node, where the limit inferior of allocation starts from kernel's end, due to lack of knowledge of memory hotplug info at this early time. But if later, hotplug info has been got, the limit inferior can be extend to 0. 'kexec -c' prefers to reuse this style to alloc mem at lower address, since if the reserved region is beyond 4G, then it requires extra mem (default is 16M) for swiotlb.Quoted 5 lines
I fail to understand why the availability of memory hotplug information would allow to extend the lower limit of bottom-up memblock allocations below the kernel. The memory in the physical range [0,
kernel_start) can be allocated as soon as the kernel memory is reserved.
Quoted 8 lines
Yes, the [0,
kernel_start) can be allocated at this time by some func e.g.memblock_reserve(). But there is trick. For the func likememblock_find_in_range(), this is hotplug attr checking ,,it will check the hotmovable attr in__next_mem_range()
{
if (movable_node_is_enabled()&&memblock_is_hotpluggable(m)) continue
}. So the movable memory can be safely skipped.
I still don't see the connection between allocating memory below kernel_start and the hotplug info.
The check for 'end > kernel_end' in
if (memblock_bottom_up() && end > kernel_end)
does not protect against allocation in a hotplugable area.
If memblock_find_in_range() is called before hotplug info is parsed it can return a range in a hotplugable area.
The point I'd like to clarify is why allocating memory in the range [0, kernel_start) cannot be done before hotplug info is available and why it is safe to allocate that memory afterwards?
Quoted 159 lines
Thanks for your kindly review.
Regards,
PingfanQuoted 153 lines
The extents of the memory node hosting the kernel image can be used to limit memblok allocations from that particular node, even in top-down mode.
Quoted 143 lines
Signed-off-by Pingfan Liu <[email protected]>Cc Tang Chen <[email protected]>Cc "Rafael J. Wysocki" <[email protected]>Cc Len Brown <[email protected]>Cc Andrew Morton <[email protected]>Cc Mike Rapoport <[email protected]>Cc Michal Hocko <[email protected]>Cc Jonathan Corbet <[email protected]>Cc Yaowei Bai <[email protected]>Cc Pavel Tatashin <[email protected]>Cc Nicholas Piggin <[email protected]>Cc Naoya Horiguchi <[email protected]>Cc Daniel Vacek <[email protected]>Cc Mathieu Malaterre <[email protected]>Cc Stefan Agner <[email protected]>Cc Dave Young <[email protected]>Cc Baoquan He <[email protected]>--- drivers/acpi/numa.c | 4 ++++ include/linux/memblock.h | 1 + mm/memblock.c | 58 +++++++++++++++++++++++++++++------------------- 3 files changed, 40 insertions(+), 23 deletions(-)drivers/acpi/numa.c +4
-0@@ -462,6 +462,10 @@ int __init acpi_numa_init(void) cnt =acpi_table_parse_srat (ACPI_SRAT_TYPE_MEMORY_AFFINITY , acpi_parse_memory_affinity,0 ); + +#if defined(CONFIG_X86 ) || defined(CONFIG_ARM64 ) +mark_mem_hotplug_parsed (); +#endif }/* SLIT: System Locality Information Table */ include/linux/memblock.h +1
-0@@ -125,6 +125,7 @@ int memblock_reserve(phys_addr_t base, phys_addr_t size);void memblock_trim_memory (phys_addr_t align);bool memblock_overlaps_region (struct memblock_type *type,phys_addr_t base,phys_addr_t size); +void mark_mem_hotplug_parsed (void );int memblock_mark_hotplug (phys_addr_t base,phys_addr_t size);int memblock_clear_hotplug (phys_addr_t base,phys_addr_t size);int memblock_mark_mirror (phys_addr_t base,phys_addr_t size);mm/memblock.c +35
-24@@ -231,6 +231,12 @@ __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,return 0 ; } +static bool mem_hotmovable_parsed __initdata_memblock; +void __init_memblock mark_mem_hotplug_parsed(void) +{ + mem_hotmovable_parsed = true; +} +/** * memblock_find_in_range_node - find free area in given range and node * @size: size of free area to find @@ -259,7 +265,7 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, phys_addr_t end, int nid, enum memblock_flags flags) { - phys_addr_t kernel_end, ret; + phys_addr_t kernel_end, ret = 0;/* pump up @end */ if (end ==MEMBLOCK_ALLOC_ACCESSIBLE )@@ -270,34 +276,40 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, end =max (start, end); kernel_end =__pa_symbol (_end); -/* -* try bottom-up allocation only when bottom-up mode -* is set and @end is above the kernel image. -*/ -if (memblock_bottom_up () && end > kernel_end) { -phys_addr_t bottom_up_start; +if (memblock_bottom_up ()) { +phys_addr_t bottom_up_start = start; -/* make sure we will allocate above the kernel */ - bottom_up_start =max (start, kernel_end); - -/* ok, try bottom-up allocation first */ - ret =__memblock_find_range_bottom_up (bottom_up_start, end, - size, align, nid, flags); -if (ret) +if (mem_hotmovable_parsed) { + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags);return ret;/* -* we always limit bottom-up allocation above the kernel, -* but top-down allocation doesn't have the limit, so -* retrying top-down allocation may succeed when bottom-up -* allocation failed. -* -* bottom-up allocation is expected to be fail very rarely, -* so we use WARN_ONCE() here to see the stack trace if -* fail happens. +* if mem hotplug info is not parsed yet, try bottom-up +* allocation with @end above the kernel image. */ -WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ), + }else if (!mem_hotmovable_parsed && end > kernel_end) { +/* make sure we will allocate above the kernel */ + bottom_up_start =max (start, kernel_end); + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags); +if (ret) +return ret; +/* +* we always limit bottom-up allocation above the +* kernel, but top-down allocation doesn't have +* the limit, so retrying top-down allocation may +* succeed when bottom-up allocation failed. +* +* bottom-up allocation is expected to be fail +* very rarely, so we use WARN_ONCE() here to see +* the stack trace if fail happens. +*/ +WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ),"memblock: bottom-up allocation failed, memory hotremove may be affected\n" ); + } } return __memblock_find_range_top_down(start, end, size, align, nid, --Quoted 2 lines
2.7.4
--
Sincerely yours,
Mike.
Signature · 2 lines
Sincerely yours, Mike.
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
Quoted 33 lines · Mike Rapoport
Quoted 16 lines · Pingfan Liu
Quoted 10 lines · Mike Rapoport
Quoted 7 lines · Pingfan Liu
The bottom-up allocation style is introduced to cope with
movable_node, where the limit inferior of allocation starts from kernel's end, due to lack of knowledge of memory hotplug info at this early time. But if later, hotplug info has been got, the limit inferior can be extend to 0. 'kexec -c' prefers to reuse this style to alloc mem at lower address, since if the reserved region is beyond 4G, then it requires extra mem (default is 16M) for swiotlb.Quoted 5 lines
I fail to understand why the availability of memory hotplug information would allow to extend the lower limit of bottom-up memblock allocations below the kernel. The memory in the physical range [0,
kernel_start) can be allocated as soon as the kernel memory is reserved.Quoted 8 lines
Yes, the [0,
kernel_start) can be allocated at this time by some func e.g.memblock_reserve(). But there is trick. For the func likememblock_find_in_range(), this is hotplug attr checking ,,it will check the hotmovable attr in__next_mem_range()
{
if (movable_node_is_enabled()&&memblock_is_hotpluggable(m)) continue
}. So the movable memory can be safely skipped.I still don't see the connection between allocating memory below
kernel_startand the hotplug info.The check for 'end >
kernel_end' inif (
memblock_bottom_up()&& end >kernel_end)
Quoted 7 lines
does not protect against allocation in a hotplugable area.
Ifmemblock_find_in_range()is called before hotplug info is parsed it can return a range in a hotplugable area.The point I'd like to clarify is why allocating memory in the range [0,
kernel_start) cannot be done before hotplug info is available and why it is safe to allocate that memory afterwards?
Well, I think that's because we have KASLR. Before KASLR was introdueced, kernel is put at a low and fixed physical address. Allocating memblock bottom-up after kernel can make sure those kernel data is in the same node as kernel text itself before SRAT parsed. While [0, kernel_start) is a very small range, e.g on x86, it was 16 MB, which is very possibly used up.
But now, with KASLR enabled by default, this bottom-up after kernel text allocation has potential issue. E.g we have node0 (including normal zone), node1(including movable zone), if KASLR put kernel at top of node0, the next memblock allocation before SRAT parsed will stamp into movable zone of node1, hotplug doesn't work well any more consequently. I had considered this issue previously, but haven't thought of a way to fix it.
While it's not related to this patch. About this patchset, I didn't check it carefully in v2 post, and acked it. In fact the current way is not good, Pingfan should call __memblock_find_range_bottom_up() directly for crashkernel reserving. Reasons are:
1)SRAT parsing is done, system restore to take top-down way to do memblock allocat.
2)we do need to find range bottom-up if user specify crashkernel=xxM (without a explicit base address).
Thanks
Baoquan
Quoted 165 lines
Quoted 160 lines
Thanks for your kindly review.
Regards,
PingfanQuoted 153 lines
The extents of the memory node hosting the kernel image can be used to limit memblok allocations from that particular node, even in top-down mode.
Quoted 143 lines
Signed-off-by Pingfan Liu <[email protected]>Cc Tang Chen <[email protected]>Cc "Rafael J. Wysocki" <[email protected]>Cc Len Brown <[email protected]>Cc Andrew Morton <[email protected]>Cc Mike Rapoport <[email protected]>Cc Michal Hocko <[email protected]>Cc Jonathan Corbet <[email protected]>Cc Yaowei Bai <[email protected]>Cc Pavel Tatashin <[email protected]>Cc Nicholas Piggin <[email protected]>Cc Naoya Horiguchi <[email protected]>Cc Daniel Vacek <[email protected]>Cc Mathieu Malaterre <[email protected]>Cc Stefan Agner <[email protected]>Cc Dave Young <[email protected]>Cc Baoquan He <[email protected]>--- drivers/acpi/numa.c | 4 ++++ include/linux/memblock.h | 1 + mm/memblock.c | 58 +++++++++++++++++++++++++++++------------------- 3 files changed, 40 insertions(+), 23 deletions(-)drivers/acpi/numa.c +4
-0@@ -462,6 +462,10 @@ int __init acpi_numa_init(void) cnt =acpi_table_parse_srat (ACPI_SRAT_TYPE_MEMORY_AFFINITY , acpi_parse_memory_affinity,0 ); + +#if defined(CONFIG_X86 ) || defined(CONFIG_ARM64 ) +mark_mem_hotplug_parsed (); +#endif }/* SLIT: System Locality Information Table */ include/linux/memblock.h +1
-0@@ -125,6 +125,7 @@ int memblock_reserve(phys_addr_t base, phys_addr_t size);void memblock_trim_memory (phys_addr_t align);bool memblock_overlaps_region (struct memblock_type *type,phys_addr_t base,phys_addr_t size); +void mark_mem_hotplug_parsed (void );int memblock_mark_hotplug (phys_addr_t base,phys_addr_t size);int memblock_clear_hotplug (phys_addr_t base,phys_addr_t size);int memblock_mark_mirror (phys_addr_t base,phys_addr_t size);mm/memblock.c +35
-24@@ -231,6 +231,12 @@ __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,return 0 ; } +static bool mem_hotmovable_parsed __initdata_memblock; +void __init_memblock mark_mem_hotplug_parsed(void) +{ + mem_hotmovable_parsed = true; +} +/** * memblock_find_in_range_node - find free area in given range and node * @size: size of free area to find @@ -259,7 +265,7 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, phys_addr_t end, int nid, enum memblock_flags flags) { - phys_addr_t kernel_end, ret; + phys_addr_t kernel_end, ret = 0;/* pump up @end */ if (end ==MEMBLOCK_ALLOC_ACCESSIBLE )@@ -270,34 +276,40 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, end =max (start, end); kernel_end =__pa_symbol (_end); -/* -* try bottom-up allocation only when bottom-up mode -* is set and @end is above the kernel image. -*/ -if (memblock_bottom_up () && end > kernel_end) { -phys_addr_t bottom_up_start; +if (memblock_bottom_up ()) { +phys_addr_t bottom_up_start = start; -/* make sure we will allocate above the kernel */ - bottom_up_start =max (start, kernel_end); - -/* ok, try bottom-up allocation first */ - ret =__memblock_find_range_bottom_up (bottom_up_start, end, - size, align, nid, flags); -if (ret) +if (mem_hotmovable_parsed) { + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags);return ret;/* -* we always limit bottom-up allocation above the kernel, -* but top-down allocation doesn't have the limit, so -* retrying top-down allocation may succeed when bottom-up -* allocation failed. -* -* bottom-up allocation is expected to be fail very rarely, -* so we use WARN_ONCE() here to see the stack trace if -* fail happens. +* if mem hotplug info is not parsed yet, try bottom-up +* allocation with @end above the kernel image. */ -WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ), + }else if (!mem_hotmovable_parsed && end > kernel_end) { +/* make sure we will allocate above the kernel */ + bottom_up_start =max (start, kernel_end); + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags); +if (ret) +return ret; +/* +* we always limit bottom-up allocation above the +* kernel, but top-down allocation doesn't have +* the limit, so retrying top-down allocation may +* succeed when bottom-up allocation failed. +* +* bottom-up allocation is expected to be fail +* very rarely, so we use WARN_ONCE() here to see +* the stack trace if fail happens. +*/ +WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ),"memblock: bottom-up allocation failed, memory hotremove may be affected\n" ); + } } return __memblock_find_range_top_down(start, end, size, align, nid, --Quoted 2 lines
2.7.4
--
Sincerely yours,
Mike.Signature · 2 lines
Sincerely yours, Mike.
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
(added Tejun)
Quoted 49 lines · Baoquan He
Quoted 33 lines · Mike Rapoport
Quoted 16 lines · Pingfan Liu
Quoted 10 lines · Mike Rapoport
Quoted 7 lines · Pingfan Liu
The bottom-up allocation style is introduced to cope with
movable_node, where the limit inferior of allocation starts from kernel's end, due to lack of knowledge of memory hotplug info at this early time. But if later, hotplug info has been got, the limit inferior can be extend to 0. 'kexec -c' prefers to reuse this style to alloc mem at lower address, since if the reserved region is beyond 4G, then it requires extra mem (default is 16M) for swiotlb.Quoted 5 lines
I fail to understand why the availability of memory hotplug information would allow to extend the lower limit of bottom-up memblock allocations below the kernel. The memory in the physical range [0,
kernel_start) can be allocated as soon as the kernel memory is reserved.Quoted 8 lines
Yes, the [0,
kernel_start) can be allocated at this time by some func e.g.memblock_reserve(). But there is trick. For the func likememblock_find_in_range(), this is hotplug attr checking ,,it will check the hotmovable attr in__next_mem_range()
{
if (movable_node_is_enabled()&&memblock_is_hotpluggable(m)) continue
}. So the movable memory can be safely skipped.I still don't see the connection between allocating memory below
kernel_startand the hotplug info.The check for 'end >
kernel_end' inif (
memblock_bottom_up()&& end >kernel_end)Quoted 7 lines
does not protect against allocation in a hotplugable area.
Ifmemblock_find_in_range()is called before hotplug info is parsed it can return a range in a hotplugable area.The point I'd like to clarify is why allocating memory in the range [0,
kernel_start) cannot be done before hotplug info is available and why it is safe to allocate that memory afterwards?Well, I think that's because we have KASLR. Before KASLR was introdueced, kernel is put at a low and fixed physical address. Allocating memblock bottom-up after kernel can make sure those kernel data is in the same node as kernel text itself before SRAT parsed. While [0,
kernel_start) is a very small range, e.g on x86, it was 16 MB, which is very possibly used up.
Quoted 7 lines
But now, with KASLR enabled by default, this bottom-up after kernel text allocation has potential issue. E.g we have node0 (including normal zone), node1(including movable zone), if KASLR put kernel at top of node0, the next memblock allocation before SRAT parsed will stamp into movable zone of node1, hotplug doesn't work well any more consequently. I had considered this issue previously, but haven't thought of a way to fix it.
I agree that currently the bottom-up allocation after the kernel text has issues with KASLR. But this issues are not necessarily related to the memory hotplug. Even with a single memory node, a bottom-up allocation will fail if KASLR would put the kernel near the end of node0.
What I am trying to understand is whether there is a fundamental reason to prevent allocations from [0, kernel_start)?
Maybe Tejun can recall why he suggested to start bottom-up allocations from kernel_end.
Quoted 178 lines
While it's not related to this patch. About this patchset, I didn't check it carefully in v2 post, and acked it. In fact the current way is not good, Pingfan should call
__memblock_find_range_bottom_up()directly for crashkernel reserving. Reasons are:
1)SRAT parsing is done, system restore to take top-down way to do memblock allocat.
2)we do need to find range bottom-up if user specify crashkernel=xxM (without a explicit base address).Thanks
BaoquanQuoted 165 lines
Quoted 160 lines
Thanks for your kindly review.
Regards,
PingfanQuoted 153 lines
The extents of the memory node hosting the kernel image can be used to limit memblok allocations from that particular node, even in top-down mode.
Quoted 143 lines
Signed-off-by Pingfan Liu <[email protected]>Cc Tang Chen <[email protected]>Cc "Rafael J. Wysocki" <[email protected]>Cc Len Brown <[email protected]>Cc Andrew Morton <[email protected]>Cc Mike Rapoport <[email protected]>Cc Michal Hocko <[email protected]>Cc Jonathan Corbet <[email protected]>Cc Yaowei Bai <[email protected]>Cc Pavel Tatashin <[email protected]>Cc Nicholas Piggin <[email protected]>Cc Naoya Horiguchi <[email protected]>Cc Daniel Vacek <[email protected]>Cc Mathieu Malaterre <[email protected]>Cc Stefan Agner <[email protected]>Cc Dave Young <[email protected]>Cc Baoquan He <[email protected]>--- drivers/acpi/numa.c | 4 ++++ include/linux/memblock.h | 1 + mm/memblock.c | 58 +++++++++++++++++++++++++++++------------------- 3 files changed, 40 insertions(+), 23 deletions(-)drivers/acpi/numa.c +4
-0@@ -462,6 +462,10 @@ int __init acpi_numa_init(void) cnt =acpi_table_parse_srat (ACPI_SRAT_TYPE_MEMORY_AFFINITY , acpi_parse_memory_affinity,0 ); + +#if defined(CONFIG_X86 ) || defined(CONFIG_ARM64 ) +mark_mem_hotplug_parsed (); +#endif }/* SLIT: System Locality Information Table */ include/linux/memblock.h +1
-0@@ -125,6 +125,7 @@ int memblock_reserve(phys_addr_t base, phys_addr_t size);void memblock_trim_memory (phys_addr_t align);bool memblock_overlaps_region (struct memblock_type *type,phys_addr_t base,phys_addr_t size); +void mark_mem_hotplug_parsed (void );int memblock_mark_hotplug (phys_addr_t base,phys_addr_t size);int memblock_clear_hotplug (phys_addr_t base,phys_addr_t size);int memblock_mark_mirror (phys_addr_t base,phys_addr_t size);mm/memblock.c +35
-24@@ -231,6 +231,12 @@ __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,return 0 ; } +static bool mem_hotmovable_parsed __initdata_memblock; +void __init_memblock mark_mem_hotplug_parsed(void) +{ + mem_hotmovable_parsed = true; +} +/** * memblock_find_in_range_node - find free area in given range and node * @size: size of free area to find @@ -259,7 +265,7 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, phys_addr_t end, int nid, enum memblock_flags flags) { - phys_addr_t kernel_end, ret; + phys_addr_t kernel_end, ret = 0;/* pump up @end */ if (end ==MEMBLOCK_ALLOC_ACCESSIBLE )@@ -270,34 +276,40 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, end =max (start, end); kernel_end =__pa_symbol (_end); -/* -* try bottom-up allocation only when bottom-up mode -* is set and @end is above the kernel image. -*/ -if (memblock_bottom_up () && end > kernel_end) { -phys_addr_t bottom_up_start; +if (memblock_bottom_up ()) { +phys_addr_t bottom_up_start = start; -/* make sure we will allocate above the kernel */ - bottom_up_start =max (start, kernel_end); - -/* ok, try bottom-up allocation first */ - ret =__memblock_find_range_bottom_up (bottom_up_start, end, - size, align, nid, flags); -if (ret) +if (mem_hotmovable_parsed) { + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags);return ret;/* -* we always limit bottom-up allocation above the kernel, -* but top-down allocation doesn't have the limit, so -* retrying top-down allocation may succeed when bottom-up -* allocation failed. -* -* bottom-up allocation is expected to be fail very rarely, -* so we use WARN_ONCE() here to see the stack trace if -* fail happens. +* if mem hotplug info is not parsed yet, try bottom-up +* allocation with @end above the kernel image. */ -WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ), + }else if (!mem_hotmovable_parsed && end > kernel_end) { +/* make sure we will allocate above the kernel */ + bottom_up_start =max (start, kernel_end); + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags); +if (ret) +return ret; +/* +* we always limit bottom-up allocation above the +* kernel, but top-down allocation doesn't have +* the limit, so retrying top-down allocation may +* succeed when bottom-up allocation failed. +* +* bottom-up allocation is expected to be fail +* very rarely, so we use WARN_ONCE() here to see +* the stack trace if fail happens. +*/ +WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ),"memblock: bottom-up allocation failed, memory hotremove may be affected\n" ); + } } return __memblock_find_range_top_down(start, end, size, align, nid, --Quoted 2 lines
2.7.4
--
Sincerely yours,
Mike.Signature · 2 lines
Sincerely yours, Mike.
Signature · 2 lines
Sincerely yours, Mike.
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
Hello,
Quoted 10 lines · Mike Rapoport
I agree that currently the bottom-up allocation after the kernel text has issues with KASLR. But this issues are not necessarily related to the memory hotplug. Even with a single memory node, a bottom-up allocation will fail if KASLR would put the kernel near the end of node0.
What I am trying to understand is whether there is a fundamental reason to prevent allocations from [0,
kernel_start)?Maybe Tejun can recall why he suggested to start bottom-up allocations from
kernel_end.
That's from 79442ed189ac ("mm/memblock.c: introduce bottom-up allocation mode"). I wasn't involved in that patch, so no idea why the restrictions were added, but FWIW it doesn't seem necessary to me.
Thanks.
Signature · 1 line
tejun
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
Quoted 11 lines · Tejun Heo
Hello,
Quoted 8 lines · Mike Rapoport
I agree that currently the bottom-up allocation after the kernel text has issues with KASLR. But this issues are not necessarily related to the memory hotplug. Even with a single memory node, a bottom-up allocation will fail if KASLR would put the kernel near the end of node0.
What I am trying to understand is whether there is a fundamental reason to prevent allocations from [0,
kernel_start)?
Quoted 6 lines
Quoted 2 lines
Maybe Tejun can recall why he suggested to start bottom-up allocations from
kernel_end.That's from
79442ed189ac("mm/memblock.c: introduce bottom-up allocation mode"). I wasn't involved in that patch, so no idea why the restrictions were added, but FWIW it doesn't seem necessary to me.
I should have added the reference [1] at the first place :) Thanks!
[1] https://lore.kernel.org/lkml/[email protected]/
Quoted 5 lines
Thanks.
Signature · 1 line
tejun
Signature · 2 lines
Sincerely yours, Mike.
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
Quoted 15 lines · Mike Rapoport
Quoted 14 lines · Tejun Heo
Hello,
Quoted 10 lines · Mike Rapoport
I agree that currently the bottom-up allocation after the kernel text has issues with KASLR. But this issues are not necessarily related to the memory hotplug. Even with a single memory node, a bottom-up allocation will fail if KASLR would put the kernel near the end of node0.
What I am trying to understand is whether there is a fundamental reason to prevent allocations from [0,
kernel_start)?Maybe Tejun can recall why he suggested to start bottom-up allocations from
kernel_end.
Quoted 8 lines
Quoted 3 lines
That's from
79442ed189ac("mm/memblock.c: introduce bottom-up allocation mode"). I wasn't involved in that patch, so no idea why the restrictions were added, but FWIW it doesn't seem necessary to me.I should have added the reference [1] at the first place :) Thanks!
With my understanding, we may not be able to discard the bottom-up method for the current kernel. It's related to hotplug feature when 'movable_node' kernel parameter is specified. With 'movable_node', system relies on reading hotplug information from firmware, on x86 it's acpi SRAT table. In the current system, we allocate memblock region top-down by default. However, before that hotplug information retrieving, there are several places of memblock allocating, top-down memblock allocation must break hotplug feature since it will allocate kernel data in movable zone which is usually at the end node on bare metal system.
This bottom-up way is taken on many ARCHes, it works well on system if KASLR is not enabled. Below is the searching result in the current linux kernel, we can see that all ARCHes have this mechanism, except of arm/arm64. But now only arm64/mips/x86 have KASLR.
W/o KASLR, allocating memblock region above kernle end when hotplug info is not parsed, looks very reasonable. Since kernel is usually put at lower address, e.g on x86, it's 16M. My thought is that we need do memblock allocation around kernel before hotplug info parsed. That is for system w/o KASLR, we will keep the current bottom-up way; for system with KASLR, we should allocate memblock region top-down just below kernel start.
This issue must break hotplug, just because currently bare metal system need add 'nokaslr' to disable KASLR since another bug fix is under discussion as below, so this issue is covered up.
[PATCH v14 0/5] x86/boot/KASLR: Parse ACPI table and limit KASLR to choosing immovable memory
lkml.kernel.org/r/[email protected]
[~ ]$ git grep memblock_set_bottom_up
arch/alpha/kernel/setup.c: memblock_set_bottom_up(true); arch/m68k/mm/motorola.c: memblock_set_bottom_up(true); arch/mips/kernel/setup.c: memblock_set_bottom_up(true); arch/mips/kernel/traps.c: memblock_set_bottom_up(false); arch/nds32/kernel/setup.c: memblock_set_bottom_up(true); arch/powerpc/kernel/paca.c: memblock_set_bottom_up(true); arch/powerpc/kernel/paca.c: memblock_set_bottom_up(false); arch/s390/kernel/setup.c: memblock_set_bottom_up(true); arch/s390/kernel/setup.c: memblock_set_bottom_up(false); arch/sparc/mm/init_32.c: memblock_set_bottom_up(true); arch/x86/kernel/setup.c: memblock_set_bottom_up(true); arch/x86/mm/numa.c: memblock_set_bottom_up(false);
include/linux/memblock.h:static inline void __init memblock_set_bottom_up(bool enable)
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
Quoted 25 lines · Baoquan He
Quoted 15 lines · Mike Rapoport
Quoted 14 lines · Tejun Heo
Hello,
Quoted 10 lines · Mike Rapoport
I agree that currently the bottom-up allocation after the kernel text has issues with KASLR. But this issues are not necessarily related to the memory hotplug. Even with a single memory node, a bottom-up allocation will fail if KASLR would put the kernel near the end of node0.
What I am trying to understand is whether there is a fundamental reason to prevent allocations from [0,
kernel_start)?Maybe Tejun can recall why he suggested to start bottom-up allocations from
kernel_end.Quoted 8 lines
Quoted 3 lines
That's from
79442ed189ac("mm/memblock.c: introduce bottom-up allocation mode"). I wasn't involved in that patch, so no idea why the restrictions were added, but FWIW it doesn't seem necessary to me.I should have added the reference [1] at the first place :) Thanks!
Quoted 9 lines
With my understanding, we may not be able to discard the bottom-up method for the current kernel. It's related to hotplug feature when '
movable_node' kernel parameter is specified. With 'movable_node', system relies on reading hotplug information from firmware, on x86 it's acpi SRAT table. In the current system, we allocate memblock region top-down by default. However, before that hotplug information retrieving, there are several places of memblock allocating, top-down memblock allocation must break hotplug feature since it will allocate kernel data in movable zone which is usually at the end node on bare metal system.
I do not suggest to discard the bottom-up method, I merely suggest to allow it to use [0, kernel_start).
Quoted 5 lines
This bottom-up way is taken on many ARCHes, it works well on system if KASLR is not enabled. Below is the searching result in the current linux kernel, we can see that all ARCHes have this mechanism, except of arm/arm64. But now only arm64/mips/x86 have KASLR.
Quoted 7 lines
W/o KASLR, allocating memblock region above kernle end when hotplug info is not parsed, looks very reasonable. Since kernel is usually put at lower address, e.g on x86, it's 16M. My thought is that we need do memblock allocation around kernel before hotplug info parsed. That is for system w/o KASLR, we will keep the current bottom-up way; for system with KASLR, we should allocate memblock region top-down just below kernel start.
I completely agree. I was thinking about making memblock_find_in_range_node() to do something like
if (memblock_bottom_up()) {
bottom_up_start = max(start, kernel_end);
ret = __memblock_find_range_bottom_up(bottom_up_start, end,
size, align, nid, flags);
if (ret)
return ret;
bottom_up_start = max(start, 0);
end = kernel_start;
ret = __memblock_find_range_top_down(bottom_up_start, end,
size, align, nid, flags);
if (ret)
return ret;
}Quoted 22 lines
This issue must break hotplug, just because currently bare metal system need add 'nokaslr' to disable KASLR since another bug fix is under discussion as below, so this issue is covered up.
[PATCH v14 0/5] x86/boot/KASLR: Parse ACPI table and limit KASLR to choosing immovable memory
lkml.kernel.org/r/[email protected][~ ]$ git grep
memblock_set_bottom_uparch/alpha/kernel/setup.c: memblock_set_bottom_up(true); arch/m68k/mm/motorola.c: memblock_set_bottom_up(true); arch/mips/kernel/setup.c: memblock_set_bottom_up(true); arch/mips/kernel/traps.c: memblock_set_bottom_up(false); arch/nds32/kernel/setup.c: memblock_set_bottom_up(true); arch/powerpc/kernel/paca.c: memblock_set_bottom_up(true); arch/powerpc/kernel/paca.c: memblock_set_bottom_up(false); arch/s390/kernel/setup.c: memblock_set_bottom_up(true); arch/s390/kernel/setup.c: memblock_set_bottom_up(false); arch/sparc/mm/init_32.c: memblock_set_bottom_up(true); arch/x86/kernel/setup.c: memblock_set_bottom_up(true); arch/x86/mm/numa.c: memblock_set_bottom_up(false);include/linux/memblock.h:static inline void
__initmemblock_set_bottom_up(bool enable)
Signature · 2 lines
Sincerely yours, Mike.
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
Quoted 2 lines · Mike Rapoport
I do not suggest to discard the bottom-up method, I merely suggest to allow it to use [0,
kernel_start).
Sorry for late reply.
I misunderstood it, sorry.
Quoted 13 lines
Quoted 5 lines
This bottom-up way is taken on many ARCHes, it works well on system if KASLR is not enabled. Below is the searching result in the current linux kernel, we can see that all ARCHes have this mechanism, except of arm/arm64. But now only arm64/mips/x86 have KASLR.
Quoted 7 lines
W/o KASLR, allocating memblock region above kernle end when hotplug info is not parsed, looks very reasonable. Since kernel is usually put at lower address, e.g on x86, it's 16M. My thought is that we need do memblock allocation around kernel before hotplug info parsed. That is for system w/o KASLR, we will keep the current bottom-up way; for system with KASLR, we should allocate memblock region top-down just below kernel start.
Quoted 5 lines
I completely agree. I was thinking about making
memblock_find_in_range_node()to do something likeif (memblock_bottom_up()) { bottom_up_start = max(start, kernel_end);
In this way, if start < kernel_end, it will still succeed to find a region in bottom-up way after kernel end.
I am still reading code. Just noticed Pingfan sent a RFC patchset to put SRAT parsing earlier, not sure if he has tested it in numa system with acpi. I doubt that really works.
Thanks
Baoquan
Quoted 16 lines
ret = __memblock_find_range_bottom_up(bottom_up_start, end, size, align, nid, flags); if (ret) return ret; bottom_up_start = max(start, 0); end = kernel_start; ret = __memblock_find_range_top_down(bottom_up_start, end, size, align, nid, flags); if (ret) return ret; }
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
In reply to Tejun HeoI send out a series [RFC PATCH 0/4] x86_64/mm: remove bottom-up allocation style by pushing forward the parsing of mem hotplug info (
https://lore.kernel.org/lkml/[email protected]/T/#t).
Please give comment if you are interested.
Thanks,
Pingfan
Quoted 23 lines · Tejun Heo
Hello,
Quoted 10 lines · Mike Rapoport
I agree that currently the bottom-up allocation after the kernel text has issues with KASLR. But this issues are not necessarily related to the memory hotplug. Even with a single memory node, a bottom-up allocation will fail if KASLR would put the kernel near the end of node0.
What I am trying to understand is whether there is a fundamental reason to prevent allocations from [0,
kernel_start)?Maybe Tejun can recall why he suggested to start bottom-up allocations from
kernel_end.That's from
79442ed189ac("mm/memblock.c: introduce bottom-up allocation mode"). I wasn't involved in that patch, so no idea why the restrictions were added, but FWIW it doesn't seem necessary to me.Thanks.
--
tejun
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
In reply to Baoquan HeQuoted 58 lines · Baoquan He
Quoted 33 lines · Mike Rapoport
Quoted 16 lines · Pingfan Liu
Quoted 10 lines · Mike Rapoport
Quoted 7 lines · Pingfan Liu
The bottom-up allocation style is introduced to cope with
movable_node, where the limit inferior of allocation starts from kernel's end, due to lack of knowledge of memory hotplug info at this early time. But if later, hotplug info has been got, the limit inferior can be extend to 0. 'kexec -c' prefers to reuse this style to alloc mem at lower address, since if the reserved region is beyond 4G, then it requires extra mem (default is 16M) for swiotlb.Quoted 5 lines
I fail to understand why the availability of memory hotplug information would allow to extend the lower limit of bottom-up memblock allocations below the kernel. The memory in the physical range [0,
kernel_start) can be allocated as soon as the kernel memory is reserved.Quoted 8 lines
Yes, the [0,
kernel_start) can be allocated at this time by some func e.g.memblock_reserve(). But there is trick. For the func likememblock_find_in_range(), this is hotplug attr checking ,,it will check the hotmovable attr in__next_mem_range()
{
if (movable_node_is_enabled()&&memblock_is_hotpluggable(m)) continue
}. So the movable memory can be safely skipped.I still don't see the connection between allocating memory below
kernel_startand the hotplug info.The check for 'end >
kernel_end' inif (memblock_bottom_up() && end > kernel_end)Quoted 7 lines
does not protect against allocation in a hotplugable area.
Ifmemblock_find_in_range()is called before hotplug info is parsed it can return a range in a hotplugable area.The point I'd like to clarify is why allocating memory in the range [0,
kernel_start) cannot be done before hotplug info is available and why it is safe to allocate that memory afterwards?Well, I think that's because we have KASLR. Before KASLR was introdueced, kernel is put at a low and fixed physical address. Allocating memblock bottom-up after kernel can make sure those kernel data is in the same node as kernel text itself before SRAT parsed. While [0,
kernel_start) is a very small range, e.g on x86, it was 16 MB, which is very possibly used up.But now, with KASLR enabled by default, this bottom-up after kernel text allocation has potential issue. E.g we have node0 (including normal zone), node1(including movable zone), if KASLR put kernel at top of node0, the next memblock allocation before SRAT parsed will stamp into movable zone of node1, hotplug doesn't work well any more consequently. I had considered this issue previously, but haven't thought of a way to fix it.
Quoted 4 lines
While it's not related to this patch. About this patchset, I didn't check it carefully in v2 post, and acked it. In fact the current way is not good, Pingfan should call
__memblock_find_range_bottom_up()directly for crashkernel reserving. Reasons are:
Good suggestion, thanks. I will send out V4.
Regards,
Pingfan
Quoted 173 lines
1)SRAT parsing is done, system restore to take top-down way to do memblock allocat.
2)we do need to find range bottom-up if user specify crashkernel=xxM (without a explicit base address).Thanks
BaoquanQuoted 165 lines
Quoted 160 lines
Thanks for your kindly review.
Regards,
PingfanQuoted 153 lines
The extents of the memory node hosting the kernel image can be used to limit memblok allocations from that particular node, even in top-down mode.
Quoted 143 lines
Signed-off-by Pingfan Liu <[email protected]>Cc Tang Chen <[email protected]>Cc "Rafael J. Wysocki" <[email protected]>Cc Len Brown <[email protected]>Cc Andrew Morton <[email protected]>Cc Mike Rapoport <[email protected]>Cc Michal Hocko <[email protected]>Cc Jonathan Corbet <[email protected]>Cc Yaowei Bai <[email protected]>Cc Pavel Tatashin <[email protected]>Cc Nicholas Piggin <[email protected]>Cc Naoya Horiguchi <[email protected]>Cc Daniel Vacek <[email protected]>Cc Mathieu Malaterre <[email protected]>Cc Stefan Agner <[email protected]>Cc Dave Young <[email protected]>Cc Baoquan He <[email protected]>--- drivers/acpi/numa.c | 4 ++++ include/linux/memblock.h | 1 + mm/memblock.c | 58 +++++++++++++++++++++++++++++------------------- 3 files changed, 40 insertions(+), 23 deletions(-)drivers/acpi/numa.c +4
-0@@ -462,6 +462,10 @@ int __init acpi_numa_init(void) cnt =acpi_table_parse_srat (ACPI_SRAT_TYPE_MEMORY_AFFINITY , acpi_parse_memory_affinity,0 ); + +#if defined(CONFIG_X86 ) || defined(CONFIG_ARM64 ) +mark_mem_hotplug_parsed (); +#endif }/* SLIT: System Locality Information Table */ include/linux/memblock.h +1
-0@@ -125,6 +125,7 @@ int memblock_reserve(phys_addr_t base, phys_addr_t size);void memblock_trim_memory (phys_addr_t align);bool memblock_overlaps_region (struct memblock_type *type,phys_addr_t base,phys_addr_t size); +void mark_mem_hotplug_parsed (void );int memblock_mark_hotplug (phys_addr_t base,phys_addr_t size);int memblock_clear_hotplug (phys_addr_t base,phys_addr_t size);int memblock_mark_mirror (phys_addr_t base,phys_addr_t size);mm/memblock.c +35
-24@@ -231,6 +231,12 @@ __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,return 0 ; } +static bool mem_hotmovable_parsed __initdata_memblock; +void __init_memblock mark_mem_hotplug_parsed(void) +{ + mem_hotmovable_parsed = true; +} +/** * memblock_find_in_range_node - find free area in given range and node * @size: size of free area to find @@ -259,7 +265,7 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, phys_addr_t end, int nid, enum memblock_flags flags) { - phys_addr_t kernel_end, ret; + phys_addr_t kernel_end, ret = 0;/* pump up @end */ if (end ==MEMBLOCK_ALLOC_ACCESSIBLE )@@ -270,34 +276,40 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, end =max (start, end); kernel_end =__pa_symbol (_end); -/* -* try bottom-up allocation only when bottom-up mode -* is set and @end is above the kernel image. -*/ -if (memblock_bottom_up () && end > kernel_end) { -phys_addr_t bottom_up_start; +if (memblock_bottom_up ()) { +phys_addr_t bottom_up_start = start; -/* make sure we will allocate above the kernel */ - bottom_up_start =max (start, kernel_end); - -/* ok, try bottom-up allocation first */ - ret =__memblock_find_range_bottom_up (bottom_up_start, end, - size, align, nid, flags); -if (ret) +if (mem_hotmovable_parsed) { + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags);return ret;/* -* we always limit bottom-up allocation above the kernel, -* but top-down allocation doesn't have the limit, so -* retrying top-down allocation may succeed when bottom-up -* allocation failed. -* -* bottom-up allocation is expected to be fail very rarely, -* so we use WARN_ONCE() here to see the stack trace if -* fail happens. +* if mem hotplug info is not parsed yet, try bottom-up +* allocation with @end above the kernel image. */ -WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ), + }else if (!mem_hotmovable_parsed && end > kernel_end) { +/* make sure we will allocate above the kernel */ + bottom_up_start =max (start, kernel_end); + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags); +if (ret) +return ret; +/* +* we always limit bottom-up allocation above the +* kernel, but top-down allocation doesn't have +* the limit, so retrying top-down allocation may +* succeed when bottom-up allocation failed. +* +* bottom-up allocation is expected to be fail +* very rarely, so we use WARN_ONCE() here to see +* the stack trace if fail happens. +*/ +WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ),"memblock: bottom-up allocation failed, memory hotremove may be affected\n" ); + } } return __memblock_find_range_top_down(start, end, size, align, nid, --Quoted 2 lines
2.7.4
--
Sincerely yours,
Mike.--
Sincerely yours,
Mike.
Re: [PATCHv3 1/2] mm/memblock: extend the limit inferior of bottom-up after parsing hotplug attr
Quoted 59 lines · Pingfan Liu
Quoted 58 lines · Baoquan He
Quoted 33 lines · Mike Rapoport
Quoted 16 lines · Pingfan Liu
Quoted 10 lines · Mike Rapoport
Quoted 7 lines · Pingfan Liu
The bottom-up allocation style is introduced to cope with
movable_node, where the limit inferior of allocation starts from kernel's end, due to lack of knowledge of memory hotplug info at this early time. But if later, hotplug info has been got, the limit inferior can be extend to 0. 'kexec -c' prefers to reuse this style to alloc mem at lower address, since if the reserved region is beyond 4G, then it requires extra mem (default is 16M) for swiotlb.Quoted 5 lines
I fail to understand why the availability of memory hotplug information would allow to extend the lower limit of bottom-up memblock allocations below the kernel. The memory in the physical range [0,
kernel_start) can be allocated as soon as the kernel memory is reserved.Quoted 8 lines
Yes, the [0,
kernel_start) can be allocated at this time by some func e.g.memblock_reserve(). But there is trick. For the func likememblock_find_in_range(), this is hotplug attr checking ,,it will check the hotmovable attr in__next_mem_range()
{
if (movable_node_is_enabled()&&memblock_is_hotpluggable(m)) continue
}. So the movable memory can be safely skipped.I still don't see the connection between allocating memory below
kernel_startand the hotplug info.The check for 'end >
kernel_end' inif (memblock_bottom_up() && end > kernel_end)Quoted 7 lines
does not protect against allocation in a hotplugable area.
Ifmemblock_find_in_range()is called before hotplug info is parsed it can return a range in a hotplugable area.The point I'd like to clarify is why allocating memory in the range [0,
kernel_start) cannot be done before hotplug info is available and why it is safe to allocate that memory afterwards?Well, I think that's because we have KASLR. Before KASLR was introdueced, kernel is put at a low and fixed physical address. Allocating memblock bottom-up after kernel can make sure those kernel data is in the same node as kernel text itself before SRAT parsed. While [0,
kernel_start) is a very small range, e.g on x86, it was 16 MB, which is very possibly used up.But now, with KASLR enabled by default, this bottom-up after kernel text allocation has potential issue. E.g we have node0 (including normal zone), node1(including movable zone), if KASLR put kernel at top of node0, the next memblock allocation before SRAT parsed will stamp into movable zone of node1, hotplug doesn't work well any more consequently. I had considered this issue previously, but haven't thought of a way to fix it.
Quoted 6 lines
Quoted 4 lines
While it's not related to this patch. About this patchset, I didn't check it carefully in v2 post, and acked it. In fact the current way is not good, Pingfan should call
__memblock_find_range_bottom_up()directly for crashkernel reserving. Reasons are:Good suggestion, thanks. I will send out V4.
I think we can simply remove the restriction of allocating above the kernel in the memblock_find_in_range_node().
Quoted 176 lines
Regards,
PingfanQuoted 173 lines
1)SRAT parsing is done, system restore to take top-down way to do memblock allocat.
2)we do need to find range bottom-up if user specify crashkernel=xxM (without a explicit base address).Thanks
BaoquanQuoted 165 lines
Quoted 160 lines
Thanks for your kindly review.
Regards,
PingfanQuoted 153 lines
The extents of the memory node hosting the kernel image can be used to limit memblok allocations from that particular node, even in top-down mode.
Quoted 143 lines
Signed-off-by Pingfan Liu <[email protected]>Cc Tang Chen <[email protected]>Cc "Rafael J. Wysocki" <[email protected]>Cc Len Brown <[email protected]>Cc Andrew Morton <[email protected]>Cc Mike Rapoport <[email protected]>Cc Michal Hocko <[email protected]>Cc Jonathan Corbet <[email protected]>Cc Yaowei Bai <[email protected]>Cc Pavel Tatashin <[email protected]>Cc Nicholas Piggin <[email protected]>Cc Naoya Horiguchi <[email protected]>Cc Daniel Vacek <[email protected]>Cc Mathieu Malaterre <[email protected]>Cc Stefan Agner <[email protected]>Cc Dave Young <[email protected]>Cc Baoquan He <[email protected]>--- drivers/acpi/numa.c | 4 ++++ include/linux/memblock.h | 1 + mm/memblock.c | 58 +++++++++++++++++++++++++++++------------------- 3 files changed, 40 insertions(+), 23 deletions(-)drivers/acpi/numa.c +4
-0@@ -462,6 +462,10 @@ int __init acpi_numa_init(void) cnt =acpi_table_parse_srat (ACPI_SRAT_TYPE_MEMORY_AFFINITY , acpi_parse_memory_affinity,0 ); + +#if defined(CONFIG_X86 ) || defined(CONFIG_ARM64 ) +mark_mem_hotplug_parsed (); +#endif }/* SLIT: System Locality Information Table */ include/linux/memblock.h +1
-0@@ -125,6 +125,7 @@ int memblock_reserve(phys_addr_t base, phys_addr_t size);void memblock_trim_memory (phys_addr_t align);bool memblock_overlaps_region (struct memblock_type *type,phys_addr_t base,phys_addr_t size); +void mark_mem_hotplug_parsed (void );int memblock_mark_hotplug (phys_addr_t base,phys_addr_t size);int memblock_clear_hotplug (phys_addr_t base,phys_addr_t size);int memblock_mark_mirror (phys_addr_t base,phys_addr_t size);mm/memblock.c +35
-24@@ -231,6 +231,12 @@ __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,return 0 ; } +static bool mem_hotmovable_parsed __initdata_memblock; +void __init_memblock mark_mem_hotplug_parsed(void) +{ + mem_hotmovable_parsed = true; +} +/** * memblock_find_in_range_node - find free area in given range and node * @size: size of free area to find @@ -259,7 +265,7 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, phys_addr_t end, int nid, enum memblock_flags flags) { - phys_addr_t kernel_end, ret; + phys_addr_t kernel_end, ret = 0;/* pump up @end */ if (end ==MEMBLOCK_ALLOC_ACCESSIBLE )@@ -270,34 +276,40 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, end =max (start, end); kernel_end =__pa_symbol (_end); -/* -* try bottom-up allocation only when bottom-up mode -* is set and @end is above the kernel image. -*/ -if (memblock_bottom_up () && end > kernel_end) { -phys_addr_t bottom_up_start; +if (memblock_bottom_up ()) { +phys_addr_t bottom_up_start = start; -/* make sure we will allocate above the kernel */ - bottom_up_start =max (start, kernel_end); - -/* ok, try bottom-up allocation first */ - ret =__memblock_find_range_bottom_up (bottom_up_start, end, - size, align, nid, flags); -if (ret) +if (mem_hotmovable_parsed) { + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags);return ret;/* -* we always limit bottom-up allocation above the kernel, -* but top-down allocation doesn't have the limit, so -* retrying top-down allocation may succeed when bottom-up -* allocation failed. -* -* bottom-up allocation is expected to be fail very rarely, -* so we use WARN_ONCE() here to see the stack trace if -* fail happens. +* if mem hotplug info is not parsed yet, try bottom-up +* allocation with @end above the kernel image. */ -WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ), + }else if (!mem_hotmovable_parsed && end > kernel_end) { +/* make sure we will allocate above the kernel */ + bottom_up_start =max (start, kernel_end); + ret =__memblock_find_range_bottom_up ( + bottom_up_start, end, size, align, nid, + flags); +if (ret) +return ret; +/* +* we always limit bottom-up allocation above the +* kernel, but top-down allocation doesn't have +* the limit, so retrying top-down allocation may +* succeed when bottom-up allocation failed. +* +* bottom-up allocation is expected to be fail +* very rarely, so we use WARN_ONCE() here to see +* the stack trace if fail happens. +*/ +WARN_ONCE (IS_ENABLED (CONFIG_MEMORY_HOTREMOVE ),"memblock: bottom-up allocation failed, memory hotremove may be affected\n" ); + } } return __memblock_find_range_top_down(start, end, size, align, nid, --Quoted 2 lines
2.7.4
--
Sincerely yours,
Mike.--
Sincerely yours,
Mike.
Signature · 2 lines
Sincerely yours, Mike.
PATCH v3 2/2 x86/kdump: bugfix, make the behavior of crashkernel=X consistent with kaslr
In reply to Pingfan LiuCustomer reported a bug on a high end server with many pcie devices, where kernel bootup with crashkernel=384M, and kaslr is enabled. Even though we still see much memory under 896 MB, the finding still failed intermittently. Because currently we can only find region under 896 MB, if w/0 ',high' specified. Then KASLR breaks 896 MB into several parts randomly, and crashkernel reservation need be aligned to 128 MB, that's why failure is found. It raises confusion to the end user that sometimes crashkernel=X works while sometimes fails.
If want to make it succeed, customer can change kernel option to "crashkernel=384M, high". Just this give "crashkernel=xx@yy" a very limited space to behave even though its grammer looks more generic. And we can't answer questions raised from customer that confidently:
1) why it doesn't succeed to reserve 896 MB;
2) what's wrong with memory region under 4G;
3) why I have to add ',high', I only require 384 MB, not 3840 MB.
This patch simplifies the method suggested in the mail [1]. It just goes bottom-up to find a candidate region for crashkernel. The bottom-up may be better compatible with the old reservation style, i.e. still want to get memory region from 896 MB firstly, then [896 MB, 4G], finally above 4G.
There is one trivial thing about the compatibility with old kexec-tools:
if the reserved region is above 896M, then old tool will fail to load bzImage. But without this patch, the old tool also fail since there is no memory below 896M can be reserved for crashkernel.
[1]: http://lists.infradead.org/pipermail/kexec/2017-October/019571.html
--- arch/x86/kernel/setup.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
arch/x86/kernel/setup.c +6 -3
@@ -541,15 +541,18 @@ static void __init reserve_crashkernel(void)/* 0 means: find the address automatically */ if (crash_base <=0 ) { +bool bottom_up =memblock_bottom_up (); + +memblock_set_bottom_up (true);/* * Set CRASH_ADDR_LOW_MAX upper bound for crash memory, * as old kexec-tools loads bzImage below that, unless * "crashkernel=size[KMG],high" is specified. */ crash_base =memblock_find_in_range (CRASH_ALIGN , - high ?CRASH_ADDR_HIGH_MAX - :CRASH_ADDR_LOW_MAX , - crash_size,CRASH_ALIGN ); + (max_pfn *PAGE_SIZE ), crash_size,CRASH_ALIGN ); +memblock_set_bottom_up (bottom_up); +if (!crash_base) {pr_info ("crashkernel reservation failed - No suitable area found.\n" );return ;
Signature · 1 line
2.7.4
Re: [PATCHv3 2/2] x86/kdump: bugfix, make the behavior of crashkernel=X consistent with kaslr
Quoted 64 lines · Pingfan Liu
Customer reported a bug on a high end server with many pcie devices, where kernel bootup with crashkernel=384M, and kaslr is enabled. Even though we still see much memory under 896 MB, the finding still failed intermittently. Because currently we can only find region under 896 MB, if w/0 ',high' specified. Then KASLR breaks 896 MB into several parts randomly, and crashkernel reservation need be aligned to 128 MB, that's why failure is found. It raises confusion to the end user that sometimes crashkernel=X works while sometimes fails.
If want to make it succeed, customer can change kernel option to "crashkernel=384M, high". Just this give "crashkernel=xx@yy" a very limited space to behave even though its grammer looks more generic. And we can't answer questions raised from customer that confidently:
1) why it doesn't succeed to reserve 896 MB;
2) what's wrong with memory region under 4G;
3) why I have to add ',high', I only require 384 MB, not 3840 MB.This patch simplifies the method suggested in the mail [1]. It just goes bottom-up to find a candidate region for crashkernel. The bottom-up may be better compatible with the old reservation style, i.e. still want to get memory region from 896 MB firstly, then [896 MB, 4G], finally above 4G.
There is one trivial thing about the compatibility with old kexec-tools:
if the reserved region is above 896M, then old tool will fail to load bzImage. But without this patch, the old tool also fail since there is no memory below 896M can be reserved for crashkernel.[1]: http://lists.infradead.org/pipermail/kexec/2017-October/019571.html
Signed-off-by Pingfan Liu <[email protected]>Cc Tang Chen <[email protected]>Cc "Rafael J. Wysocki" <[email protected]>Cc Len Brown <[email protected]>Cc Andrew Morton <akpm-de/[email protected]>Cc Mike Rapoport <[email protected]>Cc Michal Hocko <mhocko-IBi9RG/[email protected]>Cc Jonathan Corbet <[email protected]>Cc Yaowei Bai <baiyaowei-0p4V/sDNsUmm0O/7XYngnFaTQe2KTcn/@public.gmane.org>Cc Pavel Tatashin <[email protected]>Cc Nicholas Piggin <[email protected]>Cc Naoya Horiguchi <n-horiguchi-PaJj6Psr51x8UrSeD/[email protected]>Cc Daniel Vacek <[email protected]>Cc Mathieu Malaterre <[email protected]>Cc Stefan Agner <[email protected]>Cc Dave Young <[email protected]>Cc Baoquan He <[email protected]>--- arch/x86/kernel/setup.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)arch/x86/kernel/setup.c +3
-0@@ -541,15 +541,18 @@ static void __init reserve_crashkernel(void)/* 0 means: find the address automatically */ if (crash_base <=0 ) { +bool bottom_up =memblock_bottom_up (); + +memblock_set_bottom_up (true);/*
arch/x86/kernel/setup.c static void __init reserve_crashkernel(void) · line 555
Quoted 10 lines
arch/x86/kernel/setup.c +2
-3* Set CRASH_ADDR_LOW_MAX upper bound for crash memory, * as old kexec-tools loads bzImage below that, unless * "crashkernel=size[KMG],high" is specified. */ crash_base =memblock_find_in_range (CRASH_ALIGN , - high ?CRASH_ADDR_HIGH_MAX - :CRASH_ADDR_LOW_MAX , - crash_size,CRASH_ALIGN ); + (max_pfn *PAGE_SIZE ), crash_size,CRASH_ALIGN ); +memblock_set_bottom_up (bottom_up);
Using bottom-up does not guarantee that the allocation won't fall into a removable memory, it only makes it highly probable.
I think that the 'max_pfn * PAGE_SIZE' limit should be replaced with the end of the non-removable memory node.
Quoted 7 lines
arch/x86/kernel/setup.c +1
-0+if (!crash_base) {pr_info ("crashkernel reservation failed - No suitable area found.\n" );return;
Signature · 1 line
2.7.4
Signature · 2 lines
Sincerely yours, Mike.
Re: [PATCHv3 2/2] x86/kdump: bugfix, make the behavior of crashkernel=X consistent with kaslr
Quoted 80 lines · Mike Rapoport
Quoted 64 lines · Pingfan Liu
Customer reported a bug on a high end server with many pcie devices, where kernel bootup with crashkernel=384M, and kaslr is enabled. Even though we still see much memory under 896 MB, the finding still failed intermittently. Because currently we can only find region under 896 MB, if w/0 ',high' specified. Then KASLR breaks 896 MB into several parts randomly, and crashkernel reservation need be aligned to 128 MB, that's why failure is found. It raises confusion to the end user that sometimes crashkernel=X works while sometimes fails.
If want to make it succeed, customer can change kernel option to "crashkernel=384M, high". Just this give "crashkernel=xx@yy" a very limited space to behave even though its grammer looks more generic. And we can't answer questions raised from customer that confidently:
1) why it doesn't succeed to reserve 896 MB;
2) what's wrong with memory region under 4G;
3) why I have to add ',high', I only require 384 MB, not 3840 MB.This patch simplifies the method suggested in the mail [1]. It just goes bottom-up to find a candidate region for crashkernel. The bottom-up may be better compatible with the old reservation style, i.e. still want to get memory region from 896 MB firstly, then [896 MB, 4G], finally above 4G.
There is one trivial thing about the compatibility with old kexec-tools:
if the reserved region is above 896M, then old tool will fail to load bzImage. But without this patch, the old tool also fail since there is no memory below 896M can be reserved for crashkernel.[1]: http://lists.infradead.org/pipermail/kexec/2017-October/019571.html
Signed-off-by Pingfan Liu <[email protected]>Cc Tang Chen <[email protected]>Cc "Rafael J. Wysocki" <[email protected]>Cc Len Brown <[email protected]>Cc Andrew Morton <[email protected]>Cc Mike Rapoport <[email protected]>Cc Michal Hocko <[email protected]>Cc Jonathan Corbet <[email protected]>Cc Yaowei Bai <[email protected]>Cc Pavel Tatashin <[email protected]>Cc Nicholas Piggin <[email protected]>Cc Naoya Horiguchi <[email protected]>Cc Daniel Vacek <[email protected]>Cc Mathieu Malaterre <[email protected]>Cc Stefan Agner <[email protected]>Cc Dave Young <[email protected]>Cc Baoquan He <[email protected]>--- arch/x86/kernel/setup.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)arch/x86/kernel/setup.c +3
-0@@ -541,15 +541,18 @@ static void __init reserve_crashkernel(void)/* 0 means: find the address automatically */ if (crash_base <=0 ) { +bool bottom_up =memblock_bottom_up (); + +memblock_set_bottom_up (true);/* arch/x86/kernel/setup.c static void __init reserve_crashkernel(void) · line 555
Quoted 10 lines
arch/x86/kernel/setup.c +2
-3* Set CRASH_ADDR_LOW_MAX upper bound for crash memory, * as old kexec-tools loads bzImage below that, unless * "crashkernel=size[KMG],high" is specified. */ crash_base =memblock_find_in_range (CRASH_ALIGN , - high ?CRASH_ADDR_HIGH_MAX - :CRASH_ADDR_LOW_MAX , - crash_size,CRASH_ALIGN ); + (max_pfn *PAGE_SIZE ), crash_size,CRASH_ALIGN ); +memblock_set_bottom_up (bottom_up);Using bottom-up does not guarantee that the allocation won't fall into a removable memory, it only makes it highly probable.
Quoted 3 lines
I think that the '
max_pfn*PAGE_SIZE' limit should be replaced with the end of the non-removable memory node.
Since passing MEMBLOCK_NONE, memblock_find_in_range() ->...-> __next_mem_range(), there is a logic to guarantee hotmovable memory will not be stamped over.
if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) continue;
Thanks,
Pingfan
Quoted 12 lines
Quoted 7 lines
arch/x86/kernel/setup.c +1
-0+if (!crash_base) {pr_info ("crashkernel reservation failed - No suitable area found.\n" );return;--
2.7.4--
Sincerely yours,
Mike.
Re: [PATCHv3 2/2] x86/kdump: bugfix, make the behavior of crashkernel=X consistent with kaslr
Quoted 81 lines · Pingfan Liu
Quoted 80 lines · Mike Rapoport
Quoted 64 lines · Pingfan Liu
Customer reported a bug on a high end server with many pcie devices, where kernel bootup with crashkernel=384M, and kaslr is enabled. Even though we still see much memory under 896 MB, the finding still failed intermittently. Because currently we can only find region under 896 MB, if w/0 ',high' specified. Then KASLR breaks 896 MB into several parts randomly, and crashkernel reservation need be aligned to 128 MB, that's why failure is found. It raises confusion to the end user that sometimes crashkernel=X works while sometimes fails.
If want to make it succeed, customer can change kernel option to "crashkernel=384M, high". Just this give "crashkernel=xx@yy" a very limited space to behave even though its grammer looks more generic. And we can't answer questions raised from customer that confidently:
1) why it doesn't succeed to reserve 896 MB;
2) what's wrong with memory region under 4G;
3) why I have to add ',high', I only require 384 MB, not 3840 MB.This patch simplifies the method suggested in the mail [1]. It just goes bottom-up to find a candidate region for crashkernel. The bottom-up may be better compatible with the old reservation style, i.e. still want to get memory region from 896 MB firstly, then [896 MB, 4G], finally above 4G.
There is one trivial thing about the compatibility with old kexec-tools:
if the reserved region is above 896M, then old tool will fail to load bzImage. But without this patch, the old tool also fail since there is no memory below 896M can be reserved for crashkernel.[1]: http://lists.infradead.org/pipermail/kexec/2017-October/019571.html
Signed-off-by Pingfan Liu <[email protected]>Cc Tang Chen <[email protected]>Cc "Rafael J. Wysocki" <[email protected]>Cc Len Brown <[email protected]>Cc Andrew Morton <[email protected]>Cc Mike Rapoport <[email protected]>Cc Michal Hocko <[email protected]>Cc Jonathan Corbet <[email protected]>Cc Yaowei Bai <[email protected]>Cc Pavel Tatashin <[email protected]>Cc Nicholas Piggin <[email protected]>Cc Naoya Horiguchi <[email protected]>Cc Daniel Vacek <[email protected]>Cc Mathieu Malaterre <[email protected]>Cc Stefan Agner <[email protected]>Cc Dave Young <[email protected]>Cc Baoquan He <[email protected]>--- arch/x86/kernel/setup.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)arch/x86/kernel/setup.c +3
-0@@ -541,15 +541,18 @@ static void __init reserve_crashkernel(void)/* 0 means: find the address automatically */ if (crash_base <=0 ) { +bool bottom_up =memblock_bottom_up (); + +memblock_set_bottom_up (true);/* arch/x86/kernel/setup.c static void __init reserve_crashkernel(void) · line 555
Quoted 10 lines
arch/x86/kernel/setup.c +2
-3* Set CRASH_ADDR_LOW_MAX upper bound for crash memory, * as old kexec-tools loads bzImage below that, unless * "crashkernel=size[KMG],high" is specified. */ crash_base =memblock_find_in_range (CRASH_ALIGN , - high ?CRASH_ADDR_HIGH_MAX - :CRASH_ADDR_LOW_MAX , - crash_size,CRASH_ALIGN ); + (max_pfn *PAGE_SIZE ), crash_size,CRASH_ALIGN ); +memblock_set_bottom_up (bottom_up);Using bottom-up does not guarantee that the allocation won't fall into a removable memory, it only makes it highly probable.
Quoted 8 lines
Quoted 3 lines
I think that the '
max_pfn*PAGE_SIZE' limit should be replaced with the end of the non-removable memory node.Since passing
MEMBLOCK_NONE,memblock_find_in_range()->...->__next_mem_range(), there is a logic to guarantee hotmovable memory will not be stamped over.
if (movable_node_is_enabled()&&memblock_is_hotpluggable(m)) continue;
Thanks for the clarification, I've missed that.
Quoted 16 lines
Thanks,
PingfanQuoted 12 lines
Quoted 7 lines
arch/x86/kernel/setup.c +1
-0+if (!crash_base) {pr_info ("crashkernel reservation failed - No suitable area found.\n" );return;--
2.7.4--
Sincerely yours,
Mike.
Signature · 2 lines
Sincerely yours, Mike.
Quoted 4 lines · Pingfan Liu
The bottom-up allocation style is introduced to cope with
movable_node, where the limit inferior of allocation starts from kernel's end, due to lack of knowledge of memory hotplug info at this early time. Beside this original aim, 'kexec -c' prefers to reuse this style to alloc mem
Wondering what is 'kexec -c'.
Quoted 2 lines
at lower address, since if the reserved region is beyond 4G, then it requires extra mem (default is 16M) for swiotlb. But at this time hotplug info has been
The default is 256M, not sure if we are talking about the same thing.
low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
Quoted 37 lines
got, the limit inferior can be extend to 0, which is done by this series
Cc Tang Chen <[email protected]>Cc "Rafael J. Wysocki" <[email protected]>Cc Len Brown <[email protected]>Cc Andrew Morton <[email protected]>Cc Mike Rapoport <[email protected]>Cc Michal Hocko <[email protected]>Cc Jonathan Corbet <[email protected]>Cc Yaowei Bai <[email protected]>Cc Pavel Tatashin <[email protected]>Cc Nicholas Piggin <[email protected]>Cc Naoya Horiguchi <[email protected]>Cc Daniel Vacek <[email protected]>Cc Mathieu Malaterre <[email protected]>Cc Stefan Agner <[email protected]>Cc Dave Young <[email protected]>Cc Baoquan He <[email protected]>Pingfan Liu (2):
mm/memblock: extend the limit inferior of bottom-up after parsing
hotplug attr
x86/kdump: bugfix, make the behavior of crashkernel=X consistent with
kaslrarch/x86/kernel/setup.c | 9 +++++--- drivers/acpi/numa.c | 4 ++++ include/linux/memblock.h | 1 + mm/memblock.c | 58 +++++++++++++++++++++++++++++------------------- 4 files changed, 46 insertions(+), 26 deletions(-)Signature · 1 line
2.7.4