riscv: Fix linear mapping checks for non-contiguous memory regions
PATCH 1/1 riscv: Fix linear mapping checks for non-contiguous memory regions
The RISC-V kernel already has checks to ensure that memory which would lie outside of the linear mapping is not used. However those checks use memory_limit, which is used to implement the mem= kernel command line option (to limit the total amount of memory, not its address range). When memory is made up of two or more non-contiguous memory banks this check is incorrect.
Two changes are made here:
- add a call in setup_bootmem() to memblock_cap_memory_range() which will cause any memory which falls outside the linear mapping to be removed from the memory regions.
- remove the check in create_linear_mapping_page_table() which was intended to remove memory which is outside the liner mapping based on memory_limit, as it is no longer needed. Note a check for mapping more memory than memory_limit (to implement mem=) is unnecessary because of the existing call to memblock_enforce_memory_limit().
This issue was seen when booting on a SV39 platform with two memory banks:
0x00,80000000 1GiB
0x20,00000000 32GiB
This memory range is 158GiB from top to bottom, but the linear mapping is limited to 128GiB, so the lower block of RAM will be mapped at PAGE_OFFSET, and the upper block straddles the top of the linear mapping.
This causes the following Oops:
[ 0.000000] Linux version 6.10.0-rc2-gd3b8dd5b51dd-dirty ([email protected]) (riscv64-codasip-linux-gcc (GCC) 13.2.0, GNU ld (GNU Binutils) 2.41.0.20231213) #20 SMP Sat Jun 22 11:34:22 BST 2024 [ 0.000000] memblock_add: [0x0000000080000000-0x00000000bfffffff] early_init_dt_add_memory_arch+0x4a/0x52 [ 0.000000] memblock_add: [0x0000002000000000-0x00000027ffffffff] early_init_dt_add_memory_arch+0x4a/0x52 ... [ 0.000000] memblock_alloc_try_nid: 23724 bytes align=0x8 nid=-1 from=0x0000000000000000 max_addr=0x0000000000000000 early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] memblock_reserve: [0x00000027ffff5350-0x00000027ffffaffb] memblock_alloc_range_nid+0xb8/0x132 [ 0.000000] Unable to handle kernel paging request at virtual address fffffffe7fff5350 [ 0.000000] Oops [#1] [ 0.000000] Modules linked in: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.10.0-rc2-gd3b8dd5b51dd-dirty #20 [ 0.000000] Hardware name: codasip,a70x (DT) [ 0.000000] epc : __memset+0x8c/0x104 [ 0.000000] ra : memblock_alloc_try_nid+0x74/0x84 [ 0.000000] epc : ffffffff805e88c8 ra : ffffffff806148f6 sp : ffffffff80e03d50 [ 0.000000] gp : ffffffff80ec4158 tp : ffffffff80e0bec0 t0 : fffffffe7fff52f8 [ 0.000000] t1 : 00000027ffffb000 t2 : 5f6b636f6c626d65 s0 : ffffffff80e03d90 [ 0.000000] s1 : 0000000000005cac a0 : fffffffe7fff5350 a1 : 0000000000000000 [ 0.000000] a2 : 0000000000005cac a3 : fffffffe7fffaff8 a4 : 000000000000002c [ 0.000000] a5 : ffffffff805e88c8 a6 : 0000000000005cac a7 : 0000000000000030 [ 0.000000] s2 : fffffffe7fff5350 s3 : ffffffffffffffff s4 : 0000000000000000 [ 0.000000] s5 : ffffffff8062347e s6 : 0000000000000000 s7 : 0000000000000001 [ 0.000000] s8 : 0000000000002000 s9 : 00000000800226d0 s10: 0000000000000000 [ 0.000000] s11: 0000000000000000 t3 : ffffffff8080a928 t4 : ffffffff8080a928 [ 0.000000] t5 : ffffffff8080a928 t6 : ffffffff8080a940 [ 0.000000] status: 0000000200000100 badaddr: fffffffe7fff5350 cause: 000000000000000f [ 0.000000] [<ffffffff805e88c8>] __memset+0x8c/0x104 [ 0.000000] [<ffffffff8062349c>] early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] [<ffffffff8043e892>] __unflatten_device_tree+0x52/0x114 [ 0.000000] [<ffffffff8062441e>] unflatten_device_tree+0x9e/0xb8 [ 0.000000] [<ffffffff806046fe>] setup_arch+0xd4/0x5bc [ 0.000000] [<ffffffff806007aa>] start_kernel+0x76/0x81a [ 0.000000] Code: b823 02b2 bc23 02b2 b023 04b2 b423 04b2 b823 04b2 (bc23) 04b2 [ 0.000000] ---[ end trace 0000000000000000 ]--- [ 0.000000] Kernel panic - not syncing: Attempted to kill the idle task! [ 0.000000] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---
The problem is that memblock (unaware that some physical memory cannot be used) has allocated memory from the top of memory but which is outside the linear mapping region.
--- arch/riscv/mm/init.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)
arch/riscv/mm/init.c +11 -4
@@ -233,8 +233,6 @@ static void __init setup_bootmem(void)*/ memblock_reserve (vmlinux_start, vmlinux_end - vmlinux_start); - phys_ram_end =memblock_end_of_DRAM (); -/* * Make sure we align the start of the memory on a PMD boundary so that * at worst, we map the linear mapping with PMD mappings.
@@ -249,6 +247,16 @@ static void __init setup_bootmem(void)if (IS_ENABLED (CONFIG_64BIT ) &&IS_ENABLED (CONFIG_MMU )) kernel_map.va_pa_offset =PAGE_OFFSET - phys_ram_base; +/* +* The size of the linear page mapping may restrict the amount of +* usable RAM. +*/ +if (IS_ENABLED (CONFIG_64BIT )) { + max_mapped_addr =__pa (PAGE_OFFSET ) +KERN_VIRT_SIZE ; +memblock_cap_memory_range (phys_ram_base, + max_mapped_addr - phys_ram_base); + } +/* * Reserve physical address space that would be mapped to virtual * addresses greater than (void *)(-PAGE_SIZE) because:
@@ -265,6 +273,7 @@ static void __init setup_bootmem(void)memblock_reserve (max_mapped_addr, (phys_addr_t)-max_mapped_addr); } + phys_ram_end =memblock_end_of_DRAM (); min_low_pfn =PFN_UP (phys_ram_base); max_low_pfn = max_pfn =PFN_DOWN (phys_ram_end); high_memory = (void *)(__va (PFN_PHYS (max_low_pfn)));
@@ -1289,8 +1298,6 @@ static void __init create_linear_mapping_page_table(void)if (start <=__pa (PAGE_OFFSET ) &&__pa (PAGE_OFFSET ) < end) start =__pa (PAGE_OFFSET ); -if (end >=__pa (PAGE_OFFSET ) + memory_limit) - end =__pa (PAGE_OFFSET ) + memory_limit;create_linear_mapping_range (start, end,0 ); }
Signature · 1 line
2.43.0
Quoted 90 lines · Stuart Menefy
The RISC-V kernel already has checks to ensure that memory which would lie outside of the linear mapping is not used. However those checks use
memory_limit, which is used to implement the mem= kernel command line option (to limit the total amount of memory, not its address range). When memory is made up of two or more non-contiguous memory banks this check is incorrect.Two changes are made here:
- add a call insetup_bootmem()tomemblock_cap_memory_range()which will cause any memory which falls outside the linear mapping to be removed from the memory regions.
- remove the check increate_linear_mapping_page_table()which was intended to remove memory which is outside the liner mapping based onmemory_limit, as it is no longer needed. Note a check for mapping more memory thanmemory_limit(to implement mem=) is unnecessary because of the existing call tomemblock_enforce_memory_limit().This issue was seen when booting on a SV39 platform with two memory banks:
0x00,80000000 1GiB 0x20,00000000 32GiBThis memory range is 158GiB from top to bottom, but the linear mapping is limited to 128GiB, so the lower block of RAM will be mapped at
PAGE_OFFSET, and the upper block straddles the top of the linear mapping.This causes the following Oops:
[ 0.000000] Linux version 6.10.0-rc2-gd3b8dd5b51dd-dirty ([email protected]) (riscv64-codasip-linux-gcc (GCC) 13.2.0, GNU ld (GNU Binutils) 2.41.0.20231213) #20 SMP Sat Jun 22 11:34:22 BST 2024 [ 0.000000] memblock_add: [0x0000000080000000-0x00000000bfffffff] early_init_dt_add_memory_arch+0x4a/0x52 [ 0.000000] memblock_add: [0x0000002000000000-0x00000027ffffffff] early_init_dt_add_memory_arch+0x4a/0x52 ... [ 0.000000] memblock_alloc_try_nid: 23724 bytes align=0x8 nid=-1 from=0x0000000000000000 max_addr=0x0000000000000000 early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] memblock_reserve: [0x00000027ffff5350-0x00000027ffffaffb] memblock_alloc_range_nid+0xb8/0x132 [ 0.000000] Unable to handle kernel paging request at virtual address fffffffe7fff5350 [ 0.000000] Oops [#1] [ 0.000000] Modules linked in: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.10.0-rc2-gd3b8dd5b51dd-dirty #20 [ 0.000000] Hardware name: codasip,a70x (DT) [ 0.000000] epc : __memset+0x8c/0x104 [ 0.000000] ra : memblock_alloc_try_nid+0x74/0x84 [ 0.000000] epc : ffffffff805e88c8 ra : ffffffff806148f6 sp : ffffffff80e03d50 [ 0.000000] gp : ffffffff80ec4158 tp : ffffffff80e0bec0 t0 : fffffffe7fff52f8 [ 0.000000] t1 : 00000027ffffb000 t2 : 5f6b636f6c626d65 s0 : ffffffff80e03d90 [ 0.000000] s1 : 0000000000005cac a0 : fffffffe7fff5350 a1 : 0000000000000000 [ 0.000000] a2 : 0000000000005cac a3 : fffffffe7fffaff8 a4 : 000000000000002c [ 0.000000] a5 : ffffffff805e88c8 a6 : 0000000000005cac a7 : 0000000000000030 [ 0.000000] s2 : fffffffe7fff5350 s3 : ffffffffffffffff s4 : 0000000000000000 [ 0.000000] s5 : ffffffff8062347e s6 : 0000000000000000 s7 : 0000000000000001 [ 0.000000] s8 : 0000000000002000 s9 : 00000000800226d0 s10: 0000000000000000 [ 0.000000] s11: 0000000000000000 t3 : ffffffff8080a928 t4 : ffffffff8080a928 [ 0.000000] t5 : ffffffff8080a928 t6 : ffffffff8080a940 [ 0.000000] status: 0000000200000100 badaddr: fffffffe7fff5350 cause: 000000000000000f [ 0.000000] [<ffffffff805e88c8>] __memset+0x8c/0x104 [ 0.000000] [<ffffffff8062349c>] early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] [<ffffffff8043e892>] __unflatten_device_tree+0x52/0x114 [ 0.000000] [<ffffffff8062441e>] unflatten_device_tree+0x9e/0xb8 [ 0.000000] [<ffffffff806046fe>] setup_arch+0xd4/0x5bc [ 0.000000] [<ffffffff806007aa>] start_kernel+0x76/0x81a [ 0.000000] Code: b823 02b2 bc23 02b2 b023 04b2 b423 04b2 b823 04b2 (bc23) 04b2 [ 0.000000] ---[ end trace 0000000000000000 ]--- [ 0.000000] Kernel panic - not syncing: Attempted to kill the idle task! [ 0.000000] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---The problem is that memblock (unaware that some physical memory cannot be used) has allocated memory from the top of memory but which is outside the linear mapping region.
Signed-off-by Stuart Menefy <[email protected]>Fixes c99127c45248 ("riscv: Make sure the linear mapping does not use the kernel mapping")Reviewed-by David McKay <[email protected]>--- arch/riscv/mm/init.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)arch/riscv/mm/init.c +0
-2@@ -233,8 +233,6 @@ static void __init setup_bootmem(void)*/ memblock_reserve (vmlinux_start, vmlinux_end - vmlinux_start); - phys_ram_end =memblock_end_of_DRAM (); -/* * Make sure we align the start of the memory on a PMD boundary so that * at worst, we map the linear mapping with PMD mappings.
arch/riscv/mm/init.c static void __init setup_bootmem(void) · line 255
Quoted 9 lines
arch/riscv/mm/init.c +6
-0@@ -249,6 +247,16 @@ static void __init setup_bootmem(void)if (IS_ENABLED (CONFIG_64BIT ) &&IS_ENABLED (CONFIG_MMU )) kernel_map.va_pa_offset =PAGE_OFFSET - phys_ram_base; +/* +* The size of the linear page mapping may restrict the amount of +* usable RAM. +*/ +if (IS_ENABLED (CONFIG_64BIT )) { + max_mapped_addr =__pa (PAGE_OFFSET ) +KERN_VIRT_SIZE ;
This is only true for sv39 once the following patch lands https://lore.kernel.org/linux-riscv/[email protected]/
Otherwise, sv39 is "weirdly" restricted to 124GB. You mention in the changelog that the linear mapping size is limited to 128GB, does that mean you tested your patch on top this one ^? If so, would you mind adding a Tested-by to it? Otherwise, would you mind testing on top of it :) ?
I'll see with Palmer, but maybe we can't take both patches to -fixes.
You can add:
Thanks,
Alex
Quoted 24 lines
arch/riscv/mm/init.c +5
-2+memblock_cap_memory_range (phys_ram_base, + max_mapped_addr - phys_ram_base); + } +/* * Reserve physical address space that would be mapped to virtual * addresses greater than (void *)(-PAGE_SIZE) because: @@ -265,6 +273,7 @@ static void __init setup_bootmem(void)memblock_reserve (max_mapped_addr, (phys_addr_t)-max_mapped_addr); } + phys_ram_end =memblock_end_of_DRAM (); min_low_pfn =PFN_UP (phys_ram_base); max_low_pfn = max_pfn =PFN_DOWN (phys_ram_end); high_memory = (void *)(__va (PFN_PHYS (max_low_pfn)));@@ -1289,8 +1298,6 @@ static void __init create_linear_mapping_page_table(void)if (start <=__pa (PAGE_OFFSET ) &&__pa (PAGE_OFFSET ) < end) start =__pa (PAGE_OFFSET ); -if (end >=__pa (PAGE_OFFSET ) + memory_limit) - end =__pa (PAGE_OFFSET ) + memory_limit;create_linear_mapping_range (start, end,0 ); }
Quoted 96 lines · Alexandre Ghiti
Quoted 93 lines · Stuart Menefy
The RISC-V kernel already has checks to ensure that memory which would lie outside of the linear mapping is not used. However those checks use
memory_limit, which is used to implement the mem= kernel command line option (to limit the total amount of memory, not its address range). When memory is made up of two or more non-contiguous memory banks this check is incorrect.Two changes are made here:
- add a call insetup_bootmem()tomemblock_cap_memory_range()which will cause any memory which falls outside the linear mapping to be removed from the memory regions.
- remove the check increate_linear_mapping_page_table()which was intended to remove memory which is outside the liner mapping based onmemory_limit, as it is no longer needed. Note a check for mapping more memory thanmemory_limit(to implement mem=) is unnecessary because of the existing call tomemblock_enforce_memory_limit().This issue was seen when booting on a SV39 platform with two memory banks:
0x00,80000000 1GiB 0x20,00000000 32GiBThis memory range is 158GiB from top to bottom, but the linear mapping is limited to 128GiB, so the lower block of RAM will be mapped at
PAGE_OFFSET, and the upper block straddles the top of the linear mapping.This causes the following Oops:
[ 0.000000] Linux version 6.10.0-rc2-gd3b8dd5b51dd-dirty ([email protected]) (riscv64-codasip-linux-gcc (GCC) 13.2.0, GNU ld (GNU Binutils) 2.41.0.20231213) #20 SMP Sat Jun 22 11:34:22 BST 2024 [ 0.000000] memblock_add: [0x0000000080000000-0x00000000bfffffff] early_init_dt_add_memory_arch+0x4a/0x52 [ 0.000000] memblock_add: [0x0000002000000000-0x00000027ffffffff] early_init_dt_add_memory_arch+0x4a/0x52 ... [ 0.000000] memblock_alloc_try_nid: 23724 bytes align=0x8 nid=-1 from=0x0000000000000000 max_addr=0x0000000000000000 early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] memblock_reserve: [0x00000027ffff5350-0x00000027ffffaffb] memblock_alloc_range_nid+0xb8/0x132 [ 0.000000] Unable to handle kernel paging request at virtual address fffffffe7fff5350 [ 0.000000] Oops [#1] [ 0.000000] Modules linked in: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.10.0-rc2-gd3b8dd5b51dd-dirty #20 [ 0.000000] Hardware name: codasip,a70x (DT) [ 0.000000] epc : __memset+0x8c/0x104 [ 0.000000] ra : memblock_alloc_try_nid+0x74/0x84 [ 0.000000] epc : ffffffff805e88c8 ra : ffffffff806148f6 sp : ffffffff80e03d50 [ 0.000000] gp : ffffffff80ec4158 tp : ffffffff80e0bec0 t0 : fffffffe7fff52f8 [ 0.000000] t1 : 00000027ffffb000 t2 : 5f6b636f6c626d65 s0 : ffffffff80e03d90 [ 0.000000] s1 : 0000000000005cac a0 : fffffffe7fff5350 a1 : 0000000000000000 [ 0.000000] a2 : 0000000000005cac a3 : fffffffe7fffaff8 a4 : 000000000000002c [ 0.000000] a5 : ffffffff805e88c8 a6 : 0000000000005cac a7 : 0000000000000030 [ 0.000000] s2 : fffffffe7fff5350 s3 : ffffffffffffffff s4 : 0000000000000000 [ 0.000000] s5 : ffffffff8062347e s6 : 0000000000000000 s7 : 0000000000000001 [ 0.000000] s8 : 0000000000002000 s9 : 00000000800226d0 s10: 0000000000000000 [ 0.000000] s11: 0000000000000000 t3 : ffffffff8080a928 t4 : ffffffff8080a928 [ 0.000000] t5 : ffffffff8080a928 t6 : ffffffff8080a940 [ 0.000000] status: 0000000200000100 badaddr: fffffffe7fff5350 cause: 000000000000000f [ 0.000000] [<ffffffff805e88c8>] __memset+0x8c/0x104 [ 0.000000] [<ffffffff8062349c>] early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] [<ffffffff8043e892>] __unflatten_device_tree+0x52/0x114 [ 0.000000] [<ffffffff8062441e>] unflatten_device_tree+0x9e/0xb8 [ 0.000000] [<ffffffff806046fe>] setup_arch+0xd4/0x5bc [ 0.000000] [<ffffffff806007aa>] start_kernel+0x76/0x81a [ 0.000000] Code: b823 02b2 bc23 02b2 b023 04b2 b423 04b2 b823 04b2 (bc23) 04b2 [ 0.000000] ---[ end trace 0000000000000000 ]--- [ 0.000000] Kernel panic - not syncing: Attempted to kill the idle task! [ 0.000000] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---The problem is that memblock (unaware that some physical memory cannot be used) has allocated memory from the top of memory but which is outside the linear mapping region.
Signed-off-by Stuart Menefy <[email protected]>Fixes c99127c45248 ("riscv: Make sure the linear mapping does not use the kernel mapping")Reviewed-by David McKay <[email protected]>--- arch/riscv/mm/init.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)arch/riscv/mm/init.c +0
-2@@ -233,8 +233,6 @@ static void __init setup_bootmem(void)*/ memblock_reserve (vmlinux_start, vmlinux_end - vmlinux_start); - phys_ram_end =memblock_end_of_DRAM (); -/* * Make sure we align the start of the memory on a PMD boundary so that * at worst, we map the linear mapping with PMD mappings. @@ -249,6 +247,16 @@ static void __init setup_bootmem(void)if (IS_ENABLED (CONFIG_64BIT ) &&IS_ENABLED (CONFIG_MMU )) kernel_map.va_pa_offset =PAGE_OFFSET - phys_ram_base;
Quoted 10 lines
arch/riscv/mm/init.c static void __init setup_bootmem(void) · line 255
Quoted 6 lines
arch/riscv/mm/init.c +6
-0+/* +* The size of the linear page mapping may restrict the amount of +* usable RAM. +*/ +if (IS_ENABLED (CONFIG_64BIT )) { + max_mapped_addr =__pa (PAGE_OFFSET ) +KERN_VIRT_SIZE ;This is only true for sv39 once the following patch lands
https://lore.kernel.org/linux-riscv/[email protected]/
Hi Alex
I've seen this problem whether your patch has been applied or not.
Quoted 5 lines
Otherwise, sv39 is "weirdly" restricted to 124GB. You mention in the changelog that the linear mapping size is limited to 128GB, does that mean you tested your patch on top this one ^? If so, would you mind adding a Tested-by to it? Otherwise, would you mind testing on top of it :) ?
I've tested in both cases, so I'll add a Tested-by.
Note that to actually use 128GiB on Sv39 systems yet another patch is needed which I'll also post.
Quoted 1 line
I'll see with Palmer, but maybe we can't take both patches to -fixes.
All three patches address different issues, so I think it would be safe to take them in any combination. However I understand the reluctance when making changes in this area.
Thanks
Stuart
Quoted 34 lines
You can add:
Reviewed-by Alexandre Ghiti <[email protected]>Thanks,
Alex
Quoted 24 lines
arch/riscv/mm/init.c +5
-2+memblock_cap_memory_range (phys_ram_base, + max_mapped_addr - phys_ram_base); + } +/* * Reserve physical address space that would be mapped to virtual * addresses greater than (void *)(-PAGE_SIZE) because: @@ -265,6 +273,7 @@ static void __init setup_bootmem(void)memblock_reserve (max_mapped_addr, (phys_addr_t)-max_mapped_addr); } + phys_ram_end =memblock_end_of_DRAM (); min_low_pfn =PFN_UP (phys_ram_base); max_low_pfn = max_pfn =PFN_DOWN (phys_ram_end); high_memory = (void *)(__va (PFN_PHYS (max_low_pfn)));@@ -1289,8 +1298,6 @@ static void __init create_linear_mapping_page_table(void)if (start <=__pa (PAGE_OFFSET ) &&__pa (PAGE_OFFSET ) < end) start =__pa (PAGE_OFFSET ); -if (end >=__pa (PAGE_OFFSET ) + memory_limit) - end =__pa (PAGE_OFFSET ) + memory_limit;create_linear_mapping_range (start, end,0 ); }
Alexandre Ghiti
reviewerQuoted 105 lines · Stuart Menefy
Quoted 95 lines · Alexandre Ghiti
Quoted 93 lines · Stuart Menefy
The RISC-V kernel already has checks to ensure that memory which would lie outside of the linear mapping is not used. However those checks use
memory_limit, which is used to implement the mem= kernel command line option (to limit the total amount of memory, not its address range). When memory is made up of two or more non-contiguous memory banks this check is incorrect.Two changes are made here:
- add a call insetup_bootmem()tomemblock_cap_memory_range()which will cause any memory which falls outside the linear mapping to be removed from the memory regions.
- remove the check increate_linear_mapping_page_table()which was intended to remove memory which is outside the liner mapping based onmemory_limit, as it is no longer needed. Note a check for mapping more memory thanmemory_limit(to implement mem=) is unnecessary because of the existing call tomemblock_enforce_memory_limit().This issue was seen when booting on a SV39 platform with two memory banks:
0x00,80000000 1GiB 0x20,00000000 32GiBThis memory range is 158GiB from top to bottom, but the linear mapping is limited to 128GiB, so the lower block of RAM will be mapped at
PAGE_OFFSET, and the upper block straddles the top of the linear mapping.This causes the following Oops:
[ 0.000000] Linux version 6.10.0-rc2-gd3b8dd5b51dd-dirty ([email protected]) (riscv64-codasip-linux-gcc (GCC) 13.2.0, GNU ld (GNU Binutils) 2.41.0.20231213) #20 SMP Sat Jun 22 11:34:22 BST 2024 [ 0.000000] memblock_add: [0x0000000080000000-0x00000000bfffffff] early_init_dt_add_memory_arch+0x4a/0x52 [ 0.000000] memblock_add: [0x0000002000000000-0x00000027ffffffff] early_init_dt_add_memory_arch+0x4a/0x52 ... [ 0.000000] memblock_alloc_try_nid: 23724 bytes align=0x8 nid=-1 from=0x0000000000000000 max_addr=0x0000000000000000 early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] memblock_reserve: [0x00000027ffff5350-0x00000027ffffaffb] memblock_alloc_range_nid+0xb8/0x132 [ 0.000000] Unable to handle kernel paging request at virtual address fffffffe7fff5350 [ 0.000000] Oops [#1] [ 0.000000] Modules linked in: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.10.0-rc2-gd3b8dd5b51dd-dirty #20 [ 0.000000] Hardware name: codasip,a70x (DT) [ 0.000000] epc : __memset+0x8c/0x104 [ 0.000000] ra : memblock_alloc_try_nid+0x74/0x84 [ 0.000000] epc : ffffffff805e88c8 ra : ffffffff806148f6 sp : ffffffff80e03d50 [ 0.000000] gp : ffffffff80ec4158 tp : ffffffff80e0bec0 t0 : fffffffe7fff52f8 [ 0.000000] t1 : 00000027ffffb000 t2 : 5f6b636f6c626d65 s0 : ffffffff80e03d90 [ 0.000000] s1 : 0000000000005cac a0 : fffffffe7fff5350 a1 : 0000000000000000 [ 0.000000] a2 : 0000000000005cac a3 : fffffffe7fffaff8 a4 : 000000000000002c [ 0.000000] a5 : ffffffff805e88c8 a6 : 0000000000005cac a7 : 0000000000000030 [ 0.000000] s2 : fffffffe7fff5350 s3 : ffffffffffffffff s4 : 0000000000000000 [ 0.000000] s5 : ffffffff8062347e s6 : 0000000000000000 s7 : 0000000000000001 [ 0.000000] s8 : 0000000000002000 s9 : 00000000800226d0 s10: 0000000000000000 [ 0.000000] s11: 0000000000000000 t3 : ffffffff8080a928 t4 : ffffffff8080a928 [ 0.000000] t5 : ffffffff8080a928 t6 : ffffffff8080a940 [ 0.000000] status: 0000000200000100 badaddr: fffffffe7fff5350 cause: 000000000000000f [ 0.000000] [<ffffffff805e88c8>] __memset+0x8c/0x104 [ 0.000000] [<ffffffff8062349c>] early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] [<ffffffff8043e892>] __unflatten_device_tree+0x52/0x114 [ 0.000000] [<ffffffff8062441e>] unflatten_device_tree+0x9e/0xb8 [ 0.000000] [<ffffffff806046fe>] setup_arch+0xd4/0x5bc [ 0.000000] [<ffffffff806007aa>] start_kernel+0x76/0x81a [ 0.000000] Code: b823 02b2 bc23 02b2 b023 04b2 b423 04b2 b823 04b2 (bc23) 04b2 [ 0.000000] ---[ end trace 0000000000000000 ]--- [ 0.000000] Kernel panic - not syncing: Attempted to kill the idle task! [ 0.000000] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---The problem is that memblock (unaware that some physical memory cannot be used) has allocated memory from the top of memory but which is outside the linear mapping region.
Signed-off-by Stuart Menefy <[email protected]>Fixes c99127c45248 ("riscv: Make sure the linear mapping does not use the kernel mapping")Reviewed-by David McKay <[email protected]>--- arch/riscv/mm/init.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)arch/riscv/mm/init.c +0
-2@@ -233,8 +233,6 @@ static void __init setup_bootmem(void)*/ memblock_reserve (vmlinux_start, vmlinux_end - vmlinux_start); - phys_ram_end =memblock_end_of_DRAM (); -/* * Make sure we align the start of the memory on a PMD boundary so that * at worst, we map the linear mapping with PMD mappings. @@ -249,6 +247,16 @@ static void __init setup_bootmem(void)if (IS_ENABLED (CONFIG_64BIT ) &&IS_ENABLED (CONFIG_MMU )) kernel_map.va_pa_offset =PAGE_OFFSET - phys_ram_base;Quoted 9 lines
arch/riscv/mm/init.c static void __init setup_bootmem(void) · line 255
Quoted 6 lines
arch/riscv/mm/init.c +6
-0+/* +* The size of the linear page mapping may restrict the amount of +* usable RAM. +*/ +if (IS_ENABLED (CONFIG_64BIT )) { + max_mapped_addr =__pa (PAGE_OFFSET ) +KERN_VIRT_SIZE ;This is only true for sv39 once the following patch lands
https://lore.kernel.org/linux-riscv/[email protected]/
Quoted 3 lines
Hi Alex
I've seen this problem whether your patch has been applied or not.
Sure, I just meant the use of KERN_VIRT_SIZE above, sorry it was not clear.
Quoted 7 lines
Quoted 6 lines
Otherwise, sv39 is "weirdly" restricted to 124GB. You mention in the changelog that the linear mapping size is limited to 128GB, does that mean you tested your patch on top this one ^? If so, would you mind adding a Tested-by to it? Otherwise, would you mind testing on top of it :) ?
I've tested in both cases, so I'll add a Tested-by.
Thanks!
Quoted 3 lines
Note that to actually use 128GiB on Sv39 systems yet another patch is needed which I'll also post.
Ok, I'm curious! But if indeed another patch is needed, then we need to merge the 3 at the same time. I'll add this other patch to my fixes branch.
Thanks,
Alex
Quoted 48 lines
Quoted 2 lines
I'll see with Palmer, but maybe we can't take both patches to -fixes.
All three patches address different issues, so I think it would be safe to take them in any combination. However I understand the reluctance when making changes in this area.
Thanks
Stuart
Quoted 31 lines
You can add:
Reviewed-by Alexandre Ghiti <[email protected]>Thanks,
Alex
Quoted 22 lines
arch/riscv/mm/init.c +5
-2+memblock_cap_memory_range (phys_ram_base, + max_mapped_addr - phys_ram_base); + } +/* * Reserve physical address space that would be mapped to virtual * addresses greater than (void *)(-PAGE_SIZE) because: @@ -265,6 +273,7 @@ static void __init setup_bootmem(void)memblock_reserve (max_mapped_addr, (phys_addr_t)-max_mapped_addr); } + phys_ram_end =memblock_end_of_DRAM (); min_low_pfn =PFN_UP (phys_ram_base); max_low_pfn = max_pfn =PFN_DOWN (phys_ram_end); high_memory = (void *)(__va (PFN_PHYS (max_low_pfn)));@@ -1289,8 +1298,6 @@ static void __init create_linear_mapping_page_table(void)if (start <=__pa (PAGE_OFFSET ) &&__pa (PAGE_OFFSET ) < end) start =__pa (PAGE_OFFSET ); -if (end >=__pa (PAGE_OFFSET ) + memory_limit) - end =__pa (PAGE_OFFSET ) + memory_limit;Quoted 2 lines
arch/riscv/mm/init.c static void __init create_linear_mapping_page_table(void) · line 1303
Quoted 2 lines
arch/riscv/mm/init.c +0
-0create_linear_mapping_range (start, end,0 ); }_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
Alexandre Ghiti
reviewerQuoted 133 lines · Alexandre Ghiti
Quoted 107 lines · Stuart Menefy
Quoted 95 lines · Alexandre Ghiti
Quoted 93 lines · Stuart Menefy
The RISC-V kernel already has checks to ensure that memory which would lie outside of the linear mapping is not used. However those checks use
memory_limit, which is used to implement the mem= kernel command line option (to limit the total amount of memory, not its address range). When memory is made up of two or more non-contiguous memory banks this check is incorrect.Two changes are made here:
- add a call insetup_bootmem()tomemblock_cap_memory_range()which will cause any memory which falls outside the linear mapping to be removed from the memory regions.
- remove the check increate_linear_mapping_page_table()which was intended to remove memory which is outside the liner mapping based onmemory_limit, as it is no longer needed. Note a check for mapping more memory thanmemory_limit(to implement mem=) is unnecessary because of the existing call tomemblock_enforce_memory_limit().This issue was seen when booting on a SV39 platform with two memory banks:
0x00,80000000 1GiB
0x20,00000000 32GiB
This memory range is 158GiB from top to bottom, but the linear mapping is limited to 128GiB, so the lower block of RAM will be mapped atPAGE_OFFSET, and the upper block straddles the top of the linear mapping.This causes the following Oops:
[ 0.000000] Linux version 6.10.0-rc2-gd3b8dd5b51dd-dirty ([email protected]) (riscv64-codasip-linux-gcc (GCC) 13.2.0, GNU ld (GNU Binutils) 2.41.0.20231213) #20 SMP Sat Jun 22 11:34:22 BST 2024 [ 0.000000]memblock_add: [0x0000000080000000-0x00000000bfffffff]early_init_dt_add_memory_arch+0x4a/0x52
[ 0.000000]memblock_add: [0x0000002000000000-0x00000027ffffffff]early_init_dt_add_memory_arch+0x4a/0x52
...
[ 0.000000]memblock_alloc_try_nid: 23724 bytes align=0x8 nid=-1 from=0x0000000000000000 max_addr=0x0000000000000000early_init_dt_alloc_memory_arch+0x1e/0x48
[ 0.000000]memblock_reserve: [0x00000027ffff5350-0x00000027ffffaffb]memblock_alloc_range_nid+0xb8/0x132
[ 0.000000] Unable to handle kernel paging request at virtual addressfffffffe7fff5350
[ 0.000000] Oops [#1]
[ 0.000000] Modules linked in:
[ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.10.0-rc2-gd3b8dd5b51dd-dirty #20
[ 0.000000] Hardware name: codasip,a70x (DT)
[ 0.000000] epc :__memset+0x8c/0x104
[ 0.000000] ra :memblock_alloc_try_nid+0x74/0x84
[ 0.000000] epc :ffffffff805e88c8ra :ffffffff806148f6sp :ffffffff80e03d50
[ 0.000000] gp :ffffffff80ec4158tp :ffffffff80e0bec0t0 :fffffffe7fff52f8
[ 0.000000] t1 :00000027ffffb000t2 :5f6b636f6c626d65s0 :ffffffff80e03d90
[ 0.000000] s1 :0000000000005caca0 :fffffffe7fff5350a1 : 0000000000000000
[ 0.000000] a2 :0000000000005caca3 :fffffffe7fffaff8a4 :000000000000002c
[ 0.000000] a5 :ffffffff805e88c8a6 :0000000000005caca7 : 0000000000000030
[ 0.000000] s2 :fffffffe7fff5350s3 : ffffffffffffffff s4 : 0000000000000000
[ 0.000000] s5 :ffffffff8062347es6 : 0000000000000000 s7 : 0000000000000001
[ 0.000000] s8 : 0000000000002000 s9 :00000000800226d0s10: 0000000000000000
[ 0.000000] s11: 0000000000000000 t3 :ffffffff8080a928t4 :ffffffff8080a928
[ 0.000000] t5 :ffffffff8080a928t6 :ffffffff8080a940
[ 0.000000] status: 0000000200000100 badaddr:fffffffe7fff5350cause:000000000000000f
[ 0.000000] [<ffffffff805e88c8>]__memset+0x8c/0x104
[ 0.000000] [<ffffffff8062349c>]early_init_dt_alloc_memory_arch+0x1e/0x48
[ 0.000000] [<ffffffff8043e892>]__unflatten_device_tree+0x52/0x114
[ 0.000000] [<ffffffff8062441e>]unflatten_device_tree+0x9e/0xb8
[ 0.000000] [<ffffffff806046fe>]setup_arch+0xd4/0x5bc
[ 0.000000] [<ffffffff806007aa>]start_kernel+0x76/0x81a
[ 0.000000] Code: b823 02b2 bc23 02b2 b023 04b2 b423 04b2 b823 04b2 (bc23) 04b2
[ 0.000000] ---[ end trace 0000000000000000 ]---
[ 0.000000] Kernel panic - not syncing: Attempted to kill the idle task!
[ 0.000000] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---The problem is that memblock (unaware that some physical memory cannot be used) has allocated memory from the top of memory but which is outside the linear mapping region.
Signed-off-by Stuart Menefy <[email protected]>Fixes c99127c45248 ("riscv: Make sure the linear mapping does not use the kernel mapping")Reviewed-by David McKay <[email protected]>---
arch/riscv/mm/init.c| 15 +++++++++++----1 file changed, 11 insertions(+), 4 deletions(-)arch/riscv/mm/init.c +0
-2@@ -233,8 +233,6 @@ static void __init setup_bootmem(void)*/ memblock_reserve(vmlinux_start, vmlinux_end - vmlinux_start); - phys_ram_end = memblock_end_of_DRAM(); - /** Make sure we align the start of the memory on a PMD boundary so that * at worst, we map the linear mapping with PMD mappings. @@ -249,6 +247,16 @@ static void __init setup_bootmem(void) if (IS_ENABLED(CONFIG_64BIT) && IS_ENABLED(CONFIG_MMU)) kernel_map.va_pa_offset = PAGE_OFFSET - phys_ram_base;Quoted 11 lines
Quoted 7 lines
arch/riscv/mm/init.c +2
-0+ /* +* The size of the linear page mapping may restrict the amount of
+ * usable RAM.
+ */
+ if (IS_ENABLED(CONFIG_64BIT)) {
+max_mapped_addr=__pa(PAGE_OFFSET)+KERN_VIRT_SIZE;This is only true for sv39 once the following patch lands
https://lore.kernel.org/linux-riscv/[email protected]/Quoted 3 lines
Hi Alex
I've seen this problem whether your patch has been applied or not.
Sure, I just meant the use of
KERN_VIRT_SIZEabove, sorry it was not clear.Quoted 7 lines
Quoted 6 lines
Otherwise, sv39 is "weirdly" restricted to 124GB. You mention in the changelog that the linear mapping size is limited to 128GB, does that mean you tested your patch on top this one ^? If so, would you mind adding a Tested-by to it? Otherwise, would you mind testing on top of it :) ?
I've tested in both cases, so I'll add a Tested-by.
Thanks!
Quoted 3 lines
Note that to actually use 128GiB on Sv39 systems yet another patch is needed which I'll also post.
Quoted 25 lines
Ok, I'm curious! But if indeed another patch is needed, then we need to merge the 3 at the same time. I'll add this other patch to my fixes branch.
Thanks,
Alex
Quoted 18 lines
Quoted 2 lines
I'll see with Palmer, but maybe we can't take both patches to -fixes.
All three patches address different issues, so I think it would be safe to take them in any combination. However I understand the reluctance when making changes in this area.
Thanks
Stuart
Quoted 7 lines
You can add:
Reviewed-by Alexandre Ghiti <[email protected]>Thanks,
Alex
Actually, this patch breaks KASAN on sv39:
[ 0.000000] kernel BUG at arch/riscv/mm/kasan_init.c:393!
[ 0.000000] Kernel BUG [#1]
[ 0.000000] Modules linked in:
[ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.10.0-rc5-defconfig_kasan_sparsemem_vmemmmap-g018bb5470bac #1 [ 0.000000] Hardware name: riscv-virtio,qemu (DT)
[ 0.000000] epc : kasan_shallow_populate_pud+0xf2/0x228
[ 0.000000] ra : kasan_shallow_populate_pud+0x2a/0x228
[ 0.000000] epc : ffffffff81c0d2bc ra : ffffffff81c0d1f4 sp : ffffffff82607c60
[ 0.000000] gp : ffffffff82a79548 tp : ffffffff82615740 t0 : ffffffd7fefed000
[ 0.000000] t1 : ffffffd7fefec000 t2 : 0000000000000000 s0 : ffffffff82607ce0
[ 0.000000] s1 : ffffffff82a85ef8 a0 : 000000005fbfb001 a1 : 0000400000000000
[ 0.000000] a2 : 0000000000000000 a3 : dfffffff00000000 a4 : ffffffffffffffff
[ 0.000000] a5 : 0000000000000000 a6 : ffffffd7fefecfff a7 : 1ffffffaffdfd9ff
[ 0.000000] s2 : fffffff800000000 s3 : fffffff800000000 s4 : fffffff7ffffffff
[ 0.000000] s5 : ffffffff8248ab80 s6 : 0000002000000000 s7 : 000000003fffffff
[ 0.000000] s8 : ffffffff8248ab58 s9 : ffffffff8248ab88 s10: ffffffff8248b20c
[ 0.000000] s11: ffffffff8248b210 t3 : fffffff9ffdfda00 t4 : 0000000000000200
[ 0.000000] t5 : 0000000000000000 t6 : ffffffff82ac35e0
[ 0.000000] status: 0000000200000100 badaddr: 0000000000000000 cause: 0000000000000003
[ 0.000000] [<ffffffff81c0d2bc>] kasan_shallow_populate_pud+0xf2/0x228
[ 0.000000] [<ffffffff81c0d1aa>] kasan_shallow_populate_p4d+0x108/0x128
[ 0.000000] [<ffffffff81c0d082>] kasan_shallow_populate_pgd+0x15a/0x17a
[ 0.000000] [<ffffffff81c0bf26>] kasan_init+0x178/0x384
[ 0.000000] [<ffffffff81c060c0>] setup_arch+0x90/0x10a
[ 0.000000] [<ffffffff81c005b0>] start_kernel+0x5e/0x340
[ 0.000000] Code: 6a46 6aa6 6b06 7be2 7c42 7ca2 7d02 6de2 6109 8082 (9002) 0e13
[ 0.000000] ---[ end trace 0000000000000000 ]---
[ 0.000000] Kernel panic - not syncing: Fatal exception in interrupt
[ 0.000000] ---[ end Kernel panic - not syncing: Fatal exception in interrupt ]---
I took a quick look and the issue is that KASAN is populating the kernel page table but falls onto an already populated PUD entry (which is a PGD entry on sv39) and then bugs since this should not happen.
@Stuart: Can you take a look? A simple defconfig + KASAN run on qemu with sv39 triggers the issue. I'm currently investigating another failure with rc5 and once finished I can take over/help if needed.
Thanks,
Alex
Quoted 37 lines
Quoted 26 lines
Quoted 26 lines
Quoted 26 lines
+ memblock_cap_memory_range(phys_ram_base,
+max_mapped_addr-phys_ram_base);
+ }
+
/*
* Reserve physical address space that would be mapped to virtual
* addresses greater than (void *)(-PAGE_SIZE) because:arch/riscv/mm/init.c +1
-0@@ -265,6 +273,7 @@ static void __init setup_bootmem(void) memblock_reserve(max_mapped_addr, (phys_addr_t)-max_mapped_addr); } + phys_ram_end = memblock_end_of_DRAM(); min_low_pfn = PFN_UP(phys_ram_base); max_low_pfn = max_pfn = PFN_DOWN(phys_ram_end); high_memory = (void *)(__va(PFN_PHYS(max_low_pfn)));@@ -1289,8 +1298,6 @@ static void __initcreate_linear_mapping_page_table (void)if (start <=
__pa(PAGE_OFFSET)&&__pa(PAGE_OFFSET)< end)
start =__pa(PAGE_OFFSET);
- if (end >=__pa(PAGE_OFFSET)+memory_limit) - end =__pa(PAGE_OFFSET)+memory_limit;Quoted 6 lines
Quoted 2 lines
Quoted 2 lines
create_linear_mapping_range(start, end, 0); }_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
Quoted 247 lines · Alexandre Ghiti
Quoted 168 lines · Alexandre Ghiti
Quoted 140 lines · Stuart Menefy
Quoted 128 lines · Alexandre Ghiti
Quoted 126 lines · Stuart Menefy
The RISC-V kernel already has checks to ensure that memory which would lie outside of the linear mapping is not used. However those checks use
memory_limit, which is used to implement the mem= kernel command line option (to limit the total amount of memory, not its address range). When memory is made up of two or more non-contiguous memory banks this check is incorrect.Two changes are made here:
- add a call insetup_bootmem()tomemblock_cap_memory_range()
which
will cause any memory which falls outside the linear mapping
to be
removed from the memory regions.
- remove the check increate_linear_mapping_page_table()which was intended to remove memory which is outside the liner mapping
based
onmemory_limit, as it is no longer needed. Note a check for mapping more memory thanmemory_limit(to implement mem=) is unnecessary because of the existing call tomemblock_enforce_memory_limit().This issue was seen when booting on a SV39 platform with two memory banks:
0x00,80000000 1GiB 0x20,00000000 32GiBThis memory range is 158GiB from top to bottom, but the linear mapping is limited to 128GiB, so the lower block of RAM will be mapped at
PAGE_OFFSET, and the upper block straddles the top of the linear mapping.This causes the following Oops:
[ 0.000000] Linux version 6.10.0-rc2-gd3b8dd5b51dd-dirty ([email protected]) (riscv64-codasip-linux-gcc (GCC) 13.2.0, GNU ld (GNU Binutils) 2.41.0.20231213) #20 SMP Sat Jun 22 11:34:22 BST 2024 [ 0.000000] memblock_add: [0x0000000080000000-0x00000000bfffffff] early_init_dt_add_memory_arch+0x4a/0x52 [ 0.000000] memblock_add: [0x0000002000000000-0x00000027ffffffff] early_init_dt_add_memory_arch+0x4a/0x52 ... [ 0.000000] memblock_alloc_try_nid: 23724 bytes align=0x8 nid=-1 from=0x0000000000000000 max_addr=0x0000000000000000 early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] memblock_reserve: [0x00000027ffff5350-0x00000027ffffaffb] memblock_alloc_range_nid+0xb8/0x132 [ 0.000000] Unable to handle kernel paging request at virtual address fffffffe7fff5350 [ 0.000000] Oops [#1] [ 0.000000] Modules linked in: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.10.0-rc2-gd3b8dd5b51dd-dirty #20 [ 0.000000] Hardware name: codasip,a70x (DT) [ 0.000000] epc : __memset+0x8c/0x104 [ 0.000000] ra : memblock_alloc_try_nid+0x74/0x84 [ 0.000000] epc : ffffffff805e88c8 ra : ffffffff806148f6 sp : ffffffff80e03d50 [ 0.000000] gp : ffffffff80ec4158 tp : ffffffff80e0bec0 t0 : fffffffe7fff52f8 [ 0.000000] t1 : 00000027ffffb000 t2 : 5f6b636f6c626d65 s0 : ffffffff80e03d90 [ 0.000000] s1 : 0000000000005cac a0 : fffffffe7fff5350 a1 : 0000000000000000 [ 0.000000] a2 : 0000000000005cac a3 : fffffffe7fffaff8 a4 : 000000000000002c [ 0.000000] a5 : ffffffff805e88c8 a6 : 0000000000005cac a7 : 0000000000000030 [ 0.000000] s2 : fffffffe7fff5350 s3 : ffffffffffffffff s4 : 0000000000000000 [ 0.000000] s5 : ffffffff8062347e s6 : 0000000000000000 s7 : 0000000000000001 [ 0.000000] s8 : 0000000000002000 s9 : 00000000800226d0 s10: 0000000000000000 [ 0.000000] s11: 0000000000000000 t3 : ffffffff8080a928 t4 : ffffffff8080a928 [ 0.000000] t5 : ffffffff8080a928 t6 : ffffffff8080a940 [ 0.000000] status: 0000000200000100 badaddr: fffffffe7fff5350 cause: 000000000000000f [ 0.000000] [<ffffffff805e88c8>] __memset+0x8c/0x104 [ 0.000000] [<ffffffff8062349c>] early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] [<ffffffff8043e892>] __unflatten_device_tree+0x52/0x114 [ 0.000000] [<ffffffff8062441e>] unflatten_device_tree+0x9e/0xb8 [ 0.000000] [<ffffffff806046fe>] setup_arch+0xd4/0x5bc [ 0.000000] [<ffffffff806007aa>] start_kernel+0x76/0x81a [ 0.000000] Code: b823 02b2 bc23 02b2 b023 04b2 b423 04b2 b823 04b2 (bc23) 04b2 [ 0.000000] ---[ end trace 0000000000000000 ]--- [ 0.000000] Kernel panic - not syncing: Attempted to kill the idle task! [ 0.000000] ---[ end Kernel panic - not syncing: Attempted tokill the idle task! ]---
The problem is that memblock (unaware that some physical memory cannot be used) has allocated memory from the top of memory but which is outside the linear mapping region.
Signed-off-by Stuart Menefy <[email protected]>Fixes c99127c45248 ("riscv: Make sure the linear mapping does notuse the kernel mapping")
Reviewed-by David McKay <[email protected]>--- arch/riscv/mm/init.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)arch/riscv/mm/init.c +0
-2@@ -233,8 +233,6 @@ static void __init setup_bootmem(void)*/ memblock_reserve (vmlinux_start, vmlinux_end - vmlinux_start); - phys_ram_end =memblock_end_of_DRAM (); -/* * Make sure we align the start of the memory on a PMD boundary so that
* at worst, we map the linear mapping with PMD mappings.arch/riscv/mm/init.c +0
-0@@ -249,6 +247,16 @@ static void __init setup_bootmem(void)if (IS_ENABLED (CONFIG_64BIT ) &&IS_ENABLED (CONFIG_MMU )) kernel_map.va_pa_offset =PAGE_OFFSET - phys_ram_base;Quoted 11 lines
arch/riscv/mm/init.c static void __init setup_bootmem(void) · line 255
Quoted 7 lines
arch/riscv/mm/init.c +2
-0+/* +* The size of the linear page mapping may restrict the amount of
arch/riscv/mm/init.c +4
-0+* usable RAM. +*/ +if (IS_ENABLED (CONFIG_64BIT )) { + max_mapped_addr =__pa (PAGE_OFFSET ) +KERN_VIRT_SIZE ;This is only true for sv39 once the following patch lands
https://lore.kernel.org/linux-riscv/[email protected]/Quoted 3 lines
Hi Alex
I've seen this problem whether your patch has been applied or not.
Sure, I just meant the use of
KERN_VIRT_SIZEabove, sorry it was not clear.Quoted 8 lines
Quoted 7 lines
Otherwise, sv39 is "weirdly" restricted to 124GB. You mention in the changelog that the linear mapping size is limited to 128GB, does that mean you tested your patch on top this one ^? If so, would you mind adding a Tested-by to it? Otherwise, would you mind testing on top of it
:) ?I've tested in both cases, so I'll add a Tested-by.
Thanks!
Quoted 3 lines
Note that to actually use 128GiB on Sv39 systems yet another patch is needed which I'll also post.
Quoted 28 lines
Ok, I'm curious! But if indeed another patch is needed, then we need to merge the 3 at the same time. I'll add this other patch to my fixes branch.
Thanks,
Alex
Quoted 19 lines
Quoted 2 lines
I'll see with Palmer, but maybe we can't take both patches to -fixes.
All three patches address different issues, so I think it would be safe to
take them in any combination. However I understand the reluctance when making changes in this area.Thanks
Stuart
Quoted 7 lines
You can add:
Reviewed-by Alexandre Ghiti <[email protected]>Thanks,
Alex
Actually, this patch breaks KASAN on sv39:
[ 0.000000] kernel BUG at arch/riscv/mm/kasan_init.c:393! [ 0.000000] Kernel BUG [#1] [ 0.000000] Modules linked in: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.10.0-rc5-defconfig_kasan_sparsemem_vmemmmap-g018bb5470bac #1 [ 0.000000] Hardware name: riscv-virtio,qemu (DT) [ 0.000000] epc : kasan_shallow_populate_pud+0xf2/0x228 [ 0.000000] ra : kasan_shallow_populate_pud+0x2a/0x228 [ 0.000000] epc : ffffffff81c0d2bc ra : ffffffff81c0d1f4 sp : ffffffff82607c60 [ 0.000000] gp : ffffffff82a79548 tp : ffffffff82615740 t0 : ffffffd7fefed000 [ 0.000000] t1 : ffffffd7fefec000 t2 : 0000000000000000 s0 : ffffffff82607ce0 [ 0.000000] s1 : ffffffff82a85ef8 a0 : 000000005fbfb001 a1 : 0000400000000000 [ 0.000000] a2 : 0000000000000000 a3 : dfffffff00000000 a4 : ffffffffffffffff [ 0.000000] a5 : 0000000000000000 a6 : ffffffd7fefecfff a7 : 1ffffffaffdfd9ff [ 0.000000] s2 : fffffff800000000 s3 : fffffff800000000 s4 : fffffff7ffffffff [ 0.000000] s5 : ffffffff8248ab80 s6 : 0000002000000000 s7 : 000000003fffffff [ 0.000000] s8 : ffffffff8248ab58 s9 : ffffffff8248ab88 s10: ffffffff8248b20c [ 0.000000] s11: ffffffff8248b210 t3 : fffffff9ffdfda00 t4 : 0000000000000200 [ 0.000000] t5 : 0000000000000000 t6 : ffffffff82ac35e0 [ 0.000000] status: 0000000200000100 badaddr: 0000000000000000 cause: 0000000000000003 [ 0.000000] [<ffffffff81c0d2bc>] kasan_shallow_populate_pud+0xf2/0x228 [ 0.000000] [<ffffffff81c0d1aa>] kasan_shallow_populate_p4d+0x108/0x128 [ 0.000000] [<ffffffff81c0d082>] kasan_shallow_populate_pgd+0x15a/0x17a [ 0.000000] [<ffffffff81c0bf26>] kasan_init+0x178/0x384 [ 0.000000] [<ffffffff81c060c0>] setup_arch+0x90/0x10a [ 0.000000] [<ffffffff81c005b0>] start_kernel+0x5e/0x340 [ 0.000000] Code: 6a46 6aa6 6b06 7be2 7c42 7ca2 7d02 6de2 6109 8082 (9002) 0e13 [ 0.000000] ---[ end trace 0000000000000000 ]--- [ 0.000000] Kernel panic - not syncing: Fatal exception in interrupt [ 0.000000] ---[ end Kernel panic - not syncing: Fatal exception ininterrupt ]---
Quoted 7 lines
I took a quick look and the issue is that KASAN is populating the kernel page table but falls onto an already populated PUD entry (which is a PGD entry on sv39) and then bugs since this should not happen.
@Stuart: Can you take a look? A simple defconfig + KASAN run on qemu with sv39 triggers the issue. I'm currently investigating another failure with rc5 and once finished I can take over/help if needed.
The way the current KASAN initialisation is implemented requires that the start of shallow shadow mappings are aligned on a pgdir boundary for Sv39. The address of the shadow mappings are derived from the original address shifted right by KASAN_SHADOW_SCALE_SHIFT, which is 3. So as a pgdir boundary is 1GiB this means the original address must be 8GiB aligned.
The patch to increase the size of the linear mapping from 124GiB to 128GiB moves the start address of the linear mapping by 4GiB, and as a side effect moves the start of the vmalloc region by 4GiB, thus breaking the 8GiB alignment requirement for KASAN.
I've reworked the patch to simply shift the start address of the linear mapping by 8GiB instead, and it appears to fix the KASAN problem, as well as allowing the full 128GiB to be supported. A cleaner fix might be to make the alignment requirements explicit in the definition of VMALLOC_START, or alternatively fix the KASAN initialisation to cope with non-aligned start addresses. What do you think?
Stuart
Quoted 45 lines
Thanks,
Alex
Quoted 39 lines
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
+ memblock_cap_memory_range(phys_ram_base,
+max_mapped_addr-phys_ram_base);arch/riscv/mm/init.c +2
-0+ } +/* * Reserve physical address space that would be mapped to virtual
* addresses greater than (void *)(-PAGE_SIZE) because:arch/riscv/mm/init.c +1
-0@@ -265,6 +273,7 @@ static void __init setup_bootmem(void)memblock_reserve (max_mapped_addr, (phys_addr_t)-max_mapped_addr); } + phys_ram_end =memblock_end_of_DRAM (); min_low_pfn =PFN_UP (phys_ram_base); max_low_pfn = max_pfn =PFN_DOWN (phys_ram_end);high_memory = (void *)(__va(PFN_PHYS(max_low_pfn)));arch/riscv/mm/init.c +0
-2@@ -1289,8 +1298,6 @@ static void __initcreate_linear_mapping_page_table (void ) if (start <= __pa(PAGE_OFFSET) && __pa(PAGE_OFFSET) < end) start =__pa (PAGE_OFFSET ); -if (end >=__pa (PAGE_OFFSET ) + memory_limit) - end =__pa (PAGE_OFFSET ) + memory_limit;Quoted 6 lines
Quoted 2 lines
Quoted 2 lines
arch/riscv/mm/init.c +0
-0create_linear_mapping_range (start, end,0 );}_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
This patch was applied to riscv/linux.git (fixes) by Palmer Dabbelt <[email protected]>:
Hello:
This patch was applied to riscv/linux.git (fixes)
by Palmer Dabbelt <[email protected]>:
Quoted 8 lines · On Sat, 22 Jun 2024 12:42:16 +0100 you wrote:
The RISC-V kernel already has checks to ensure that memory which would lie outside of the linear mapping is not used. However those checks use
memory_limit, which is used to implement the mem= kernel command line option (to limit the total amount of memory, not its address range). When memory is made up of two or more non-contiguous memory banks this check is incorrect.[...]
Here is the summary with links:
- riscv: Fix linear mapping checks for non-contiguous memory regions
https://git.kernel.org/riscv/c/3f4b9eeadb17
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html
Hi Stuart,
Quoted 64 lines · Stuart Menefy
The RISC-V kernel already has checks to ensure that memory which would lie outside of the linear mapping is not used. However those checks use
memory_limit, which is used to implement the mem= kernel command line option (to limit the total amount of memory, not its address range). When memory is made up of two or more non-contiguous memory banks this check is incorrect.Two changes are made here:
- add a call insetup_bootmem()tomemblock_cap_memory_range()which will cause any memory which falls outside the linear mapping to be removed from the memory regions.
- remove the check increate_linear_mapping_page_table()which was intended to remove memory which is outside the liner mapping based onmemory_limit, as it is no longer needed. Note a check for mapping more memory thanmemory_limit(to implement mem=) is unnecessary because of the existing call tomemblock_enforce_memory_limit().This issue was seen when booting on a SV39 platform with two memory banks:
0x00,80000000 1GiB
0x20,00000000 32GiB
This memory range is 158GiB from top to bottom, but the linear mapping is limited to 128GiB, so the lower block of RAM will be mapped atPAGE_OFFSET, and the upper block straddles the top of the linear mapping.This causes the following Oops:
[ 0.000000] Linux version 6.10.0-rc2-gd3b8dd5b51dd-dirty ([email protected]) (riscv64-codasip-linux-gcc (GCC) 13.2.0, GNU ld (GNU Binutils) 2.41.0.20231213) #20 SMP Sat Jun 22 11:34:22 BST 2024 [ 0.000000] memblock_add: [0x0000000080000000-0x00000000bfffffff] early_init_dt_add_memory_arch+0x4a/0x52 [ 0.000000] memblock_add: [0x0000002000000000-0x00000027ffffffff] early_init_dt_add_memory_arch+0x4a/0x52 ... [ 0.000000] memblock_alloc_try_nid: 23724 bytes align=0x8 nid=-1 from=0x0000000000000000 max_addr=0x0000000000000000 early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] memblock_reserve: [0x00000027ffff5350-0x00000027ffffaffb] memblock_alloc_range_nid+0xb8/0x132 [ 0.000000] Unable to handle kernel paging request at virtual address fffffffe7fff5350 [ 0.000000] Oops [#1] [ 0.000000] Modules linked in: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.10.0-rc2-gd3b8dd5b51dd-dirty #20 [ 0.000000] Hardware name: codasip,a70x (DT) [ 0.000000] epc : __memset+0x8c/0x104 [ 0.000000] ra : memblock_alloc_try_nid+0x74/0x84 [ 0.000000] epc : ffffffff805e88c8 ra : ffffffff806148f6 sp : ffffffff80e03d50 [ 0.000000] gp : ffffffff80ec4158 tp : ffffffff80e0bec0 t0 : fffffffe7fff52f8 [ 0.000000] t1 : 00000027ffffb000 t2 : 5f6b636f6c626d65 s0 : ffffffff80e03d90 [ 0.000000] s1 : 0000000000005cac a0 : fffffffe7fff5350 a1 : 0000000000000000 [ 0.000000] a2 : 0000000000005cac a3 : fffffffe7fffaff8 a4 : 000000000000002c [ 0.000000] a5 : ffffffff805e88c8 a6 : 0000000000005cac a7 : 0000000000000030 [ 0.000000] s2 : fffffffe7fff5350 s3 : ffffffffffffffff s4 : 0000000000000000 [ 0.000000] s5 : ffffffff8062347e s6 : 0000000000000000 s7 : 0000000000000001 [ 0.000000] s8 : 0000000000002000 s9 : 00000000800226d0 s10: 0000000000000000 [ 0.000000] s11: 0000000000000000 t3 : ffffffff8080a928 t4 : ffffffff8080a928 [ 0.000000] t5 : ffffffff8080a928 t6 : ffffffff8080a940 [ 0.000000] status: 0000000200000100 badaddr: fffffffe7fff5350 cause: 000000000000000f [ 0.000000] [<ffffffff805e88c8>] __memset+0x8c/0x104 [ 0.000000] [<ffffffff8062349c>] early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] [<ffffffff8043e892>] __unflatten_device_tree+0x52/0x114 [ 0.000000] [<ffffffff8062441e>] unflatten_device_tree+0x9e/0xb8 [ 0.000000] [<ffffffff806046fe>] setup_arch+0xd4/0x5bc [ 0.000000] [<ffffffff806007aa>] start_kernel+0x76/0x81a [ 0.000000] Code: b823 02b2 bc23 02b2 b023 04b2 b423 04b2 b823 04b2 (bc23) 04b2 [ 0.000000] ---[ end trace 0000000000000000 ]--- [ 0.000000] Kernel panic - not syncing: Attempted to kill the idle task! [ 0.000000] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---
Quoted 7 lines
The problem is that memblock (unaware that some physical memory cannot be used) has allocated memory from the top of memory but which is outside the linear mapping region.
Signed-off-by Stuart Menefy <[email protected]>Fixes c99127c45248 ("riscv: Make sure the linear mapping does not use the kernel mapping")Reviewed-by David McKay <[email protected]>
Thanks for your patch, which is now commit 3b6564427aea83b7 ("riscv:
Fix linear mapping checks for non-contiguous memory regions") in v6.11-rc2.
This causes the boot to hang on SiPEED MAiX BiT (K210 64-bit nommu):
Linux version 6.11.0-rc1-k210-00008-g3b6564427aea (geert@rox) (riscv64-linux-gnu-gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0, GNU ld (GNU Binutils for Ubuntu) 2.38) #452 SMP Tue Aug 20 15:02:49 CEST 2024 Machine model: SiPeed MAIX BiT
Forcing kernel command line to: earlycon console=ttySIF0 rootdelay=2 root=/dev/mmcblk0p1 ro
earlycon: sifive0 at MMIO 0x0000000038000000 (options '115200n8')
printk: legacy bootconsole [sifive0] enabled
Zone ranges:
DMA32 [mem 0x0000000080000000-0x00000000807fffff]
Normal empty
Movable zone start for each node
Early memory node ranges
node 0: [mem 0x0000000080000000-0x00000000807fffff]
Initmem setup node 0 [mem 0x0000000080000000-0x00000000807fffff] Falling back to deprecated "riscv,isa"
riscv: base ISA extensions acdfim
riscv: ELF capabilities acdfim
percpu: max_distance=0x1a000 too large for vmalloc space 0x0
percpu: Embedded 13 pages/cpu s20512 r0 d32736 u53248
percpu: wasting 6 pages per chunk
Kernel command line: earlycon console=ttySIF0 rootdelay=2 root=/dev/mmcblk0p1 ro Dentry cache hash table entries: 1024 (order: 1, 8192 bytes, linear) Inode-cache hash table entries: 512 (order: 0, 4096 bytes, linear) Built 1 zonelists, mobility grouping off. Total pages: 2048
mem auto-init: stack:off, heap alloc:off, heap free:off
SLUB: HWalign=64, Order=0-1, MinObjects=0, CPUs=2, Nodes=1
rcu: Hierarchical RCU implementation.
rcu: RCU calculated value of scheduler-enlistment delay is 25 jiffies.NR_IRQS: 64, nr_irqs: 64, preallocated irqs: 0
Adding a bit of debug info shows:
setup_bootmem:236: old phys_ram_end = 0x80800000
setup_bootmem:258: max_mapped_addr = 0x7fffffff
setup_bootmem:259: memblock_cap_memory_range(0x80000000, 0xffffffffffffffff)
memblock_remove_range:857: Removing region at 0x80000000 size 0x20d1e8
setup_bootmem:281: new phys_ram_end = 0x80800000Quoted 37 lines
arch/riscv/mm/init.c +11
-2@@ -233,8 +233,6 @@ static void __init setup_bootmem(void)*/ memblock_reserve (vmlinux_start, vmlinux_end - vmlinux_start); - phys_ram_end =memblock_end_of_DRAM (); -/* * Make sure we align the start of the memory on a PMD boundary so that * at worst, we map the linear mapping with PMD mappings. @@ -249,6 +247,16 @@ static void __init setup_bootmem(void)if (IS_ENABLED (CONFIG_64BIT ) &&IS_ENABLED (CONFIG_MMU )) kernel_map.va_pa_offset =PAGE_OFFSET - phys_ram_base; +/* +* The size of the linear page mapping may restrict the amount of +* usable RAM. +*/ +if (IS_ENABLED (CONFIG_64BIT )) { + max_mapped_addr =__pa (PAGE_OFFSET ) +KERN_VIRT_SIZE ; +memblock_cap_memory_range (phys_ram_base, + max_mapped_addr - phys_ram_base); + } +/* * Reserve physical address space that would be mapped to virtual * addresses greater than (void *)(-PAGE_SIZE) because: @@ -265,6 +273,7 @@ static void __init setup_bootmem(void)memblock_reserve (max_mapped_addr, (phys_addr_t)-max_mapped_addr); } + phys_ram_end =memblock_end_of_DRAM (); min_low_pfn =PFN_UP (phys_ram_base); max_low_pfn = max_pfn =PFN_DOWN (phys_ram_end); high_memory = (void *)(__va (PFN_PHYS (max_low_pfn)));
arch/riscv/mm/init.c static void __init create_linear_mapping_page_table(void) · line 1303
Quoted 8 lines
arch/riscv/mm/init.c +0
-2@@ -1289,8 +1298,6 @@ static void __init create_linear_mapping_page_table(void)if (start <=__pa (PAGE_OFFSET ) &&__pa (PAGE_OFFSET ) < end) start =__pa (PAGE_OFFSET ); -if (end >=__pa (PAGE_OFFSET ) + memory_limit) - end =__pa (PAGE_OFFSET ) + memory_limit;create_linear_mapping_range (start, end,0 ); }
Gr{oetje,eeting}s,
Geert
Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected] In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds
Alexandre Ghiti
reviewerHi Geert,
Quoted 111 lines · Geert Uytterhoeven
Hi Stuart,
Quoted 64 lines · Stuart Menefy
The RISC-V kernel already has checks to ensure that memory which would lie outside of the linear mapping is not used. However those checks use
memory_limit, which is used to implement the mem= kernel command line option (to limit the total amount of memory, not its address range). When memory is made up of two or more non-contiguous memory banks this check is incorrect.Two changes are made here:
- add a call insetup_bootmem()tomemblock_cap_memory_range()which will cause any memory which falls outside the linear mapping to be removed from the memory regions.
- remove the check increate_linear_mapping_page_table()which was intended to remove memory which is outside the liner mapping based onmemory_limit, as it is no longer needed. Note a check for mapping more memory thanmemory_limit(to implement mem=) is unnecessary because of the existing call tomemblock_enforce_memory_limit().This issue was seen when booting on a SV39 platform with two memory banks:
0x00,80000000 1GiB 0x20,00000000 32GiBThis memory range is 158GiB from top to bottom, but the linear mapping is limited to 128GiB, so the lower block of RAM will be mapped at
PAGE_OFFSET, and the upper block straddles the top of the linear mapping.This causes the following Oops:
[ 0.000000] Linux version 6.10.0-rc2-gd3b8dd5b51dd-dirty ([email protected]) (riscv64-codasip-linux-gcc (GCC) 13.2.0, GNU ld (GNU Binutils) 2.41.0.20231213) #20 SMP Sat Jun 22 11:34:22 BST 2024 [ 0.000000] memblock_add: [0x0000000080000000-0x00000000bfffffff] early_init_dt_add_memory_arch+0x4a/0x52 [ 0.000000] memblock_add: [0x0000002000000000-0x00000027ffffffff] early_init_dt_add_memory_arch+0x4a/0x52 ... [ 0.000000] memblock_alloc_try_nid: 23724 bytes align=0x8 nid=-1 from=0x0000000000000000 max_addr=0x0000000000000000 early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] memblock_reserve: [0x00000027ffff5350-0x00000027ffffaffb] memblock_alloc_range_nid+0xb8/0x132 [ 0.000000] Unable to handle kernel paging request at virtual address fffffffe7fff5350 [ 0.000000] Oops [#1] [ 0.000000] Modules linked in: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.10.0-rc2-gd3b8dd5b51dd-dirty #20 [ 0.000000] Hardware name: codasip,a70x (DT) [ 0.000000] epc : __memset+0x8c/0x104 [ 0.000000] ra : memblock_alloc_try_nid+0x74/0x84 [ 0.000000] epc : ffffffff805e88c8 ra : ffffffff806148f6 sp : ffffffff80e03d50 [ 0.000000] gp : ffffffff80ec4158 tp : ffffffff80e0bec0 t0 : fffffffe7fff52f8 [ 0.000000] t1 : 00000027ffffb000 t2 : 5f6b636f6c626d65 s0 : ffffffff80e03d90 [ 0.000000] s1 : 0000000000005cac a0 : fffffffe7fff5350 a1 : 0000000000000000 [ 0.000000] a2 : 0000000000005cac a3 : fffffffe7fffaff8 a4 : 000000000000002c [ 0.000000] a5 : ffffffff805e88c8 a6 : 0000000000005cac a7 : 0000000000000030 [ 0.000000] s2 : fffffffe7fff5350 s3 : ffffffffffffffff s4 : 0000000000000000 [ 0.000000] s5 : ffffffff8062347e s6 : 0000000000000000 s7 : 0000000000000001 [ 0.000000] s8 : 0000000000002000 s9 : 00000000800226d0 s10: 0000000000000000 [ 0.000000] s11: 0000000000000000 t3 : ffffffff8080a928 t4 : ffffffff8080a928 [ 0.000000] t5 : ffffffff8080a928 t6 : ffffffff8080a940 [ 0.000000] status: 0000000200000100 badaddr: fffffffe7fff5350 cause: 000000000000000f [ 0.000000] [<ffffffff805e88c8>] __memset+0x8c/0x104 [ 0.000000] [<ffffffff8062349c>] early_init_dt_alloc_memory_arch+0x1e/0x48 [ 0.000000] [<ffffffff8043e892>] __unflatten_device_tree+0x52/0x114 [ 0.000000] [<ffffffff8062441e>] unflatten_device_tree+0x9e/0xb8 [ 0.000000] [<ffffffff806046fe>] setup_arch+0xd4/0x5bc [ 0.000000] [<ffffffff806007aa>] start_kernel+0x76/0x81a [ 0.000000] Code: b823 02b2 bc23 02b2 b023 04b2 b423 04b2 b823 04b2 (bc23) 04b2 [ 0.000000] ---[ end trace 0000000000000000 ]--- [ 0.000000] Kernel panic - not syncing: Attempted to kill the idle task! [ 0.000000] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---Quoted 7 lines
The problem is that memblock (unaware that some physical memory cannot be used) has allocated memory from the top of memory but which is outside the linear mapping region.
Signed-off-by Stuart Menefy <[email protected]>Fixes c99127c45248 ("riscv: Make sure the linear mapping does not use the kernel mapping")Reviewed-by David McKay <[email protected]>Thanks for your patch, which is now commit
3b6564427aea83b7("riscv:
Fix linear mapping checks for non-contiguous memory regions") in v6.11-rc2.This causes the boot to hang on SiPEED MAiX BiT (K210 64-bit nommu):
Linux version 6.11.0-rc1-k210-00008-g3b6564427aea (geert@rox) (riscv64-linux-gnu-gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0, GNU ld (GNU Binutils for Ubuntu) 2.38) #452 SMP Tue Aug 20 15:02:49 CEST 2024 Machine model: SiPeed MAIX BiT
Forcing kernel command line to: earlycon console=ttySIF0 rootdelay=2 root=/dev/mmcblk0p1 ro
earlycon: sifive0 at MMIO0x0000000038000000(options '115200n8')
printk: legacy bootconsole [sifive0] enabled
Zone ranges:DMA32 [mem 0x0000000080000000-0x00000000807fffff] Normal emptyMovable zone start for each node
Early memory node ranges
node 0: [mem 0x0000000080000000-0x00000000807fffff]
Initmem setup node 0 [mem 0x0000000080000000-0x00000000807fffff] Falling back to deprecated "riscv,isa"
riscv: base ISA extensions acdfim
riscv: ELF capabilities acdfim
percpu: max_distance=0x1a000 too large for vmalloc space0x0
percpu: Embedded 13 pages/cpu s20512 r0 d32736 u53248
percpu: wasting 6 pages per chunk
Kernel command line: earlycon console=ttySIF0 rootdelay=2 root=/dev/mmcblk0p1 ro Dentry cache hash table entries: 1024 (order: 1, 8192 bytes, linear) Inode-cache hash table entries: 512 (order: 0, 4096 bytes, linear) Built 1 zonelists, mobility grouping off. Total pages: 2048
mem auto-init: stack:off, heap alloc:off, heap free:off
SLUB: HWalign=64, Order=0-1, MinObjects=0, CPUs=2, Nodes=1
rcu: Hierarchical RCU implementation.
rcu: RCU calculated value of scheduler-enlistment delay is 25 jiffies.NR_IRQS: 64,nr_irqs: 64, preallocated irqs: 0
Quoted 7 lines
Adding a bit of debug info shows:
setup_bootmem:236: old phys_ram_end = 0x80800000 setup_bootmem:258: max_mapped_addr = 0x7fffffff setup_bootmem:259: memblock_cap_memory_range(0x80000000, 0xffffffffffffffff) memblock_remove_range:857: Removing region at 0x80000000 size 0x20d1e8 setup_bootmem:281: new phys_ram_end = 0x80800000
Thank you very much for the report!
The following patch should fix this, can you give it a try?
arch/riscv/mm/init.c +0 -0
@@ -252,7 +252,7 @@ static void __init setup_bootmem(void)
* The size of the linear page mapping may restrict the amount of
* usable RAM.
*/
- if (IS_ENABLED(CONFIG_64BIT)) {
+ if (IS_ENABLED(CONFIG_64BIT) && IS_ENABLED(CONFIG_MMU)) { max_mapped_addr = __pa(PAGE_OFFSET) + KERN_VIRT_SIZE; memblock_cap_memory_range(phys_ram_base, max_mapped_addr - phys_ram_base);
It makes no sense to restrict physical memory size because of linear mapping size constraints when there is no linear mapping :)
I'll send a proper patch when you confirm it works for you :)
Thanks again,
Alex
Quoted 49 lines
Quoted 37 lines
arch/riscv/mm/init.c +11
-2@@ -233,8 +233,6 @@ static void __init setup_bootmem(void)*/ memblock_reserve (vmlinux_start, vmlinux_end - vmlinux_start); - phys_ram_end =memblock_end_of_DRAM (); -/* * Make sure we align the start of the memory on a PMD boundary so that * at worst, we map the linear mapping with PMD mappings. @@ -249,6 +247,16 @@ static void __init setup_bootmem(void)if (IS_ENABLED (CONFIG_64BIT ) &&IS_ENABLED (CONFIG_MMU )) kernel_map.va_pa_offset =PAGE_OFFSET - phys_ram_base; +/* +* The size of the linear page mapping may restrict the amount of +* usable RAM. +*/ +if (IS_ENABLED (CONFIG_64BIT )) { + max_mapped_addr =__pa (PAGE_OFFSET ) +KERN_VIRT_SIZE ; +memblock_cap_memory_range (phys_ram_base, + max_mapped_addr - phys_ram_base); + } +/* * Reserve physical address space that would be mapped to virtual * addresses greater than (void *)(-PAGE_SIZE) because: @@ -265,6 +273,7 @@ static void __init setup_bootmem(void)memblock_reserve (max_mapped_addr, (phys_addr_t)-max_mapped_addr); } + phys_ram_end =memblock_end_of_DRAM (); min_low_pfn =PFN_UP (phys_ram_base); max_low_pfn = max_pfn =PFN_DOWN (phys_ram_end); high_memory = (void *)(__va (PFN_PHYS (max_low_pfn)));arch/riscv/mm/init.c static void __init create_linear_mapping_page_table(void) · line 1303
Quoted 8 lines
arch/riscv/mm/init.c +0
-2@@ -1289,8 +1298,6 @@ static void __init create_linear_mapping_page_table(void)if (start <=__pa (PAGE_OFFSET ) &&__pa (PAGE_OFFSET ) < end) start =__pa (PAGE_OFFSET ); -if (end >=__pa (PAGE_OFFSET ) + memory_limit) - end =__pa (PAGE_OFFSET ) + memory_limit;create_linear_mapping_range (start, end,0 ); }Gr{oetje,eeting}s,
Geert