Re: [PATCH 02/02] scripts/gdb/mm: Add support for powerpc book3s64

[Ritesh Harjani (IBM) <ritesh.list@gmail.com>]

Jan Kiszka <jan.kiszka@siemens.com> writes:

> On 30.08.25 05:45, Ritesh Harjani (IBM) wrote:
>> This adds page ops support to powerpc book3s64. Following operations are
>> now supported:
>> 
>> lx-pfn_to_kaddr -- PFN to kernel address
>> lx-pfn_to_page -- PFN to struct page
>> lx-page_address -- struct page to linear mapping address
>> lx-page_to_pfn -- struct page to PFN
>> lx-page_to_phys -- struct page to physical address
>> lx-virt_to_page -- virtual address to struct page
>> lx-virt_to_phys -- virtual address to physical address
>> 
>> lx-vmallocinfo -- Show vmallocinfo
>> lx-slabinfo -- Show slabinfo
>> 
>> e.g. Below showing lx-mmu_info command i.e.
>> On Radix:
>> (gdb) lx-mmu_info
>> MMU: Radix
>> 
>> On Hash:
>> (gdb) lx-mmu_info
>> MMU: Hash
>> 
>> e.g. Below shows that struct page pointers coming from vmemmap area i.e.
>> (gdb) p vmemmap
>> $5 = (struct page *) 0xc00c000000000000
>> 
>> (gdb) lx-pfn_to_page 0
>> pfn_to_page(0x0) = 0xc00c000000000000
>> 
>> (gdb) lx-pfn_to_page 1
>> pfn_to_page(0x0) = 0xc00c000000000040
>> 
>> Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
>> ---
>>  scripts/gdb/linux/constants.py.in |   4 ++
>>  scripts/gdb/linux/mm.py           | 114 +++++++++++++++++++++++++++++-
>>  2 files changed, 116 insertions(+), 2 deletions(-)
>> 
>> diff --git a/scripts/gdb/linux/constants.py.in b/scripts/gdb/linux/constants.py.in
>> index 154db10fe94a..97a731db6e89 100644
>> --- a/scripts/gdb/linux/constants.py.in
>> +++ b/scripts/gdb/linux/constants.py.in
>> @@ -153,6 +153,10 @@ if IS_BUILTIN(CONFIG_ARM64):
>>      LX_VALUE(CONFIG_PAGE_SHIFT)
>>      LX_VALUE(CONFIG_ARCH_FORCE_MAX_ORDER)
>>  LX_CONFIG(CONFIG_PPC_BOOK3S_64)
>> +if IS_BUILTIN(CONFIG_PPC_BOOK3S_64):
>> +    LX_VALUE(CONFIG_PAGE_OFFSET)
>> +    LX_VALUE(CONFIG_PAGE_SHIFT)
>> +    LX_VALUE(CONFIG_KERNEL_START)
>>  LX_CONFIG(CONFIG_SPARSEMEM)
>>  LX_CONFIG(CONFIG_SPARSEMEM_EXTREME)
>>  LX_CONFIG(CONFIG_SPARSEMEM_VMEMMAP)
>> diff --git a/scripts/gdb/linux/mm.py b/scripts/gdb/linux/mm.py
>> index 7571aebbe650..9e5b1632f910 100644
>> --- a/scripts/gdb/linux/mm.py
>> +++ b/scripts/gdb/linux/mm.py
>> @@ -24,10 +24,17 @@ class page_ops():
>>      def __init__(self):
>>          if not constants.LX_CONFIG_SPARSEMEM_VMEMMAP:
>>              raise gdb.GdbError('Only support CONFIG_SPARSEMEM_VMEMMAP now')
>> -        if constants.LX_CONFIG_ARM64 and utils.is_target_arch('aarch64'):
>> +
>> +        if utils.is_target_arch('aarch64'):
>> +            if not constants.LX_CONFIG_ARM64:
>
> This reorders the check, pulling the dynamic part before the static one
> - why? Not that this is run on every command, but at least
> initialization could slow down by some cycles (or more...).
>

Sure that make sense Jan, 
Let me fix that up for both aarch64 and powerpc in v2.
Since it's been sometime that I have looked at these patches,
let me go over them once again before I re-spin a v2.

-ritesh

>> +                raise gdb.GdbError('ARM64 page ops require CONFIG_ARM64')
>>              self.ops = aarch64_page_ops()
>> +        elif utils.is_target_arch('powerpc'):
>> +            if not constants.LX_CONFIG_PPC_BOOK3S_64:
>> +                raise gdb.GdbError('Only supported for Book3s_64')
>> +            self.ops = powerpc64_page_ops()
>>          else:
>> -            raise gdb.GdbError('Only support aarch64 now')
>> +            raise gdb.GdbError('Unsupported arch for page ops')
>> 
>>  class aarch64_page_ops():
>>      def __init__(self):
>> @@ -287,6 +294,109 @@ class aarch64_page_ops():
>>      def folio_address(self, folio):
>>          return self.page_address(folio['page'].address)
>> 
>> +
>> +class powerpc64_page_ops():
>> +    """powerpc64 minimal Virtual Memory operations
>> +    """
>> +
>> +    def __init__(self):
>> +        vmemmap_sym = gdb.parse_and_eval('vmemmap')
>> +        self.vmemmap = vmemmap_sym.cast(utils.get_page_type().pointer())
>> +
>> +        self.PAGE_SHIFT = constants.LX_CONFIG_PAGE_SHIFT
>> +        self.PAGE_OFFSET = constants.LX_CONFIG_PAGE_OFFSET
>> +        self.KERNEL_START = constants.LX_CONFIG_KERNEL_START
>> +
>> +        # These variables are common for both Hash and Radix so no
>> +        # need to explicitely check for MMU mode.
>> +        self.KERNEL_VIRT_START = gdb.parse_and_eval("__kernel_virt_start")
>> +        self.VMALLOC_START = gdb.parse_and_eval("__vmalloc_start")
>> +        self.VMALLOC_END = gdb.parse_and_eval("__vmalloc_end")
>> +        self.KERNEL_IO_START = gdb.parse_and_eval("__kernel_io_start")
>> +        self.KERNEL_IO_END = gdb.parse_and_eval("__kernel_io_end")
>> +        # KERN_MAP_SIZE can be calculated from below trick to avoid
>> +        # checking Hash 4k/64k pagesize
>> +        self.KERN_MAP_SIZE = self.KERNEL_IO_END - self.KERNEL_IO_START
>> +        self.VMEMMAP_START = gdb.parse_and_eval("vmemmap")
>> +        self.VMEMMAP_SIZE = self.KERN_MAP_SIZE
>> +        self.VMEMMAP_END = self.VMEMMAP_START + self.VMEMMAP_SIZE
>> +
>> +        if constants.LX_CONFIG_NUMA and constants.LX_CONFIG_NODES_SHIFT:
>> +            self.NODE_SHIFT = constants.LX_CONFIG_NODES_SHIFT
>> +        else:
>> +            self.NODE_SHIFT = 0
>> +        self.MAX_NUMNODES = 1 << self.NODE_SHIFT
>> +
>> +    def PFN_PHYS(self, pfn):
>> +        return pfn << self.PAGE_SHIFT
>> +
>> +    def PHYS_PFN(self, pfn):
>> +        return pfn >> self.PAGE_SHIFT
>> +
>> +    def __va(self, pa):
>> +        return pa | self.PAGE_OFFSET
>> +
>> +    def __pa(self, va):
>> +        return va & 0x0fffffffffffffff;
>> +
>> +    def pfn_to_page(self, pfn):
>> +        return (self.vmemmap + int(pfn)).cast(utils.get_page_type().pointer())
>> +
>> +    def page_to_pfn(self, page):
>> +        pagep = page.cast(utils.get_page_type().pointer())
>> +        return int(pagep - self.vmemmap)
>> +
>> +    def page_address(self, page):
>> +        pfn = self.page_to_pfn(page)
>> +        va = self.PAGE_OFFSET + (pfn << self.PAGE_SHIFT)
>> +        return va
>> +
>> +    def page_to_phys(self, page):
>> +        pfn = self.page_to_pfn(page)
>> +        return self.PFN_PHYS(pfn)
>> +
>> +    def phys_to_page(self, pa):
>> +        pfn = self.PHYS_PFN(pa)
>> +        return self.pfn_to_page(pfn)
>> +
>> +    def phys_to_virt(self, pa):
>> +        return self.__va(pa)
>> +
>> +    def virt_to_phys(self, va):
>> +        return self.__pa(va)
>> +
>> +    def virt_to_pfn(self, va):
>> +        return self.__pa(va) >> self.PAGE_SHIFT
>> +
>> +    def virt_to_page(self, va):
>> +        return self.pfn_to_page(self.virt_to_pfn(va))
>> +
>> +    def pfn_to_kaddr(self, pfn):
>> +        return self.__va(pfn << self.PAGE_SHIFT)
>> +
>> +    # powerpc does not use tags for KASAN. So simply return addr
>> +    def kasan_reset_tag(self, addr):
>> +        return addr
>> +
>> +class LxMmuInfo(gdb.Command):
>> +    """MMU Type for PowerPC Book3s64"""
>> +
>> +    def __init__(self):
>> +        super(LxMmuInfo, self).__init__("lx-mmu_info", gdb.COMMAND_USER)
>> +
>> +    def invoke(self, arg, from_tty):
>> +        if not constants.LX_CONFIG_PPC_BOOK3S_64:
>> +            raise gdb.GdbError("Only supported for Book3s_64")
>> +
>> +        lpcr = gdb.parse_and_eval("(unsigned long)$lpcr")
>> +        # Host Radix bit should be 1 in LPCR for Radix MMU
>> +        if (lpcr & 0x0000000000100000):
>> +            gdb.write("MMU: Radix\n")
>> +        else:
>> +            gdb.write("MMU: Hash\n")
>> +
>> +LxMmuInfo()
>> +
>>  class LxPFN2Page(gdb.Command):
>>      """PFN to struct page"""
>> 
>> --
>> 2.50.1
>> 
>
> Rest (including patch 1) looks good to me, though not looking deep into
> the architectural bits.
>
> Jan
>
> PS: Sorry for the late feedback.
>
> -- 
> Siemens AG, Foundational Technologies
> Linux Expert Center