Hi Tao
attached split 2 patches.
Thanks
Guanyou
Tao Liu <ltao(a)redhat.com> 于2025年12月5日周五 11:31写道:
Hi Guanyou,
On Fri, Dec 5, 2025 at 3:03 PM Guanyou Chen <chenguanyou9338(a)gmail.com>
wrote:
>
> Hi Tao,
>
> Check swap_file_is_file from crash commit 3452fe802bf94
> ("Fix "kmem -i" and "swap" commands on Linux 6.10-rc1 and
later kernels")
>
> Thanks,
> Guanyou
>
> Tao Liu <ltao(a)redhat.com> 于2025年12月5日周五 06:40写道:
>>
>> Hi Guanyou,
>>
>> On Thu, Dec 4, 2025 at 11:33 PM Guanyou Chen <chenguanyou9338(a)gmail.com>
wrote:
>> >
>> > Hi Lianbo
>> >
>> > Kernel commit 55564814a838 ("arm64/mm: Move PTE_PRESENT_INVALID
>> > to overlay PTE_NG"), which is contained in Linux 6.10 and
>> > later kernels, changed the format of swap entries on arm64.
>> > Without the patch, the "vtop" command cannot display swap
information
>> >
>> > before:
>> > crash> vtop 2000000
>> > VIRTUAL PHYSICAL
>> > 2000000 (not mapped)
>> >
>> > PAGE DIRECTORY: ffffff8106356000
>> > PGD: ffffff8106356000 => 800000186355003
>> > PMD: ffffff8106355080 => 8000001476f5003
>> > PTE: ffffff80c76f5000 => 101a62004
>> >
>> > PTE OFFSET: 1055330
>> > vtop: cannot determine swap location
>> >
>> > without the patch:
>> > crash> vtop 2000000
>> > VIRTUAL PHYSICAL
>> > 2000000 (not mapped)
>> >
>> > PAGE DIRECTORY: ffffff8106356000
>> > PGD: ffffff8106356000 => 800000186355003
>> > PMD: ffffff8106355080 => 8000001476f5003
>> > PTE: ffffff80c76f5000 => 101a62004
>> >
>> > PTE SWAP OFFSET
>> > 101a62004 /first_stage_ramdisk/dev/block/zram0 1055330
>> >
>> > VMA START END FLAGS FILE
>> > ffffff81a06e8b00 2000000 22000000 100073
>> >
>> > SWAP: /first_stage_ramdisk/dev/block/zram0 OFFSET: 1055330
>> >
>> > Link:
https://lore.kernel.org/r/20240503144604.151095-4-ryan.roberts@arm.com
>> > Signed-off-by: Guanyou.Chen <chenguanyou(a)xiaomi.com>
>> > ---
>> > arm64.c | 11 ++++++++++-
>> > memory.c | 5 ++++-
>> > 2 files changed, 14 insertions(+), 2 deletions(-)
>> >
>> > diff --git a/arm64.c b/arm64.c
>> > index 1723595..c125655 100644
>> > --- a/arm64.c
>> > +++ b/arm64.c
>> > @@ -712,7 +712,16 @@ arm64_init(int when)
>> > }
>> > }
>> >
>> > - if (THIS_KERNEL_VERSION >= LINUX(5,19,0)) {
>> > + if (THIS_KERNEL_VERSION >= LINUX(6,10,0)) {
>> > + ms->__SWP_TYPE_BITS = 5;
>> > + ms->__SWP_TYPE_SHIFT = 6;
>> > + ms->__SWP_TYPE_MASK = ((1UL << ms->__SWP_TYPE_BITS)
- 1);
>> > + ms->__SWP_OFFSET_SHIFT = 12;
>> > + ms->__SWP_OFFSET_BITS = 50;
>> > + ms->__SWP_OFFSET_MASK = ((1UL <<
ms->__SWP_OFFSET_BITS) -
1);
>> > + ms->PTE_PROT_NONE = 0; /* unused */
>> > + ms->PTE_FILE = 0; /* unused */
>> > + } else if (THIS_KERNEL_VERSION >= LINUX(5,19,0)) {
Sorry maybe I didn't make myself clear. What I mean is, the kernel
link, aka
https://lore.kernel.org/r/20240503144604.151095-4-ryan.roberts@arm.com,
updated the information as you have addressed above, and this part is
OK to me.
>> > ms->__SWP_TYPE_BITS = 5;
>> > ms->__SWP_TYPE_SHIFT = 3;
>> > ms->__SWP_TYPE_MASK = ((1UL << ms->__SWP_TYPE_BITS)
- 1);
>> > diff --git a/memory.c b/memory.c
>> > index 400d31a..cbc8d2f 100644
>> > --- a/memory.c
>> > +++ b/memory.c
>> > @@ -16415,6 +16415,8 @@ get_swapdev(ulong type, char *buf)
>> > ulong vfsmnt;
>> > char *devname;
>> > char buf1[BUFSIZE];
>> > + int swap_file_is_file =
>> > + STREQ(MEMBER_TYPE_NAME("swap_info_struct",
"swap_file"),
"file");
>> >
>> > swap_info_init();
>> >
>> > @@ -16474,7 +16476,8 @@ get_swapdev(ulong type, char *buf)
>> > vfsmnt = ULONG(vt->swap_info_struct +
>> > OFFSET(swap_info_struct_swap_vfsmnt));
>> > get_pathname(swap_file, buf, BUFSIZE, 1, vfsmnt);
>> > - } else if (VALID_MEMBER
(swap_info_struct_old_block_size)) {
>> > + } else if (VALID_MEMBER
(swap_info_struct_old_block_size)
>> > + || swap_file_is_file) {
But the link nor your commit log doesn't provide information for the
above swap_file_is_file code change. Like why do you need to add the
"|| swap_file_is_file" check here? Does it have any relations with the
previous __SWP_TYPE_XXX value updating? What I'm trying to figure out
is whether your __SWP_TYPE_XXX updating and swap_file_is_file
checking, are independent of each other and should go into 2 seperated
patches.
>>
>> Any reason for the extra swap_file_is_file check? From the kernel
>> patch link you posted, only the __SWP_TYPE_XXX changes, which is
>> presented in (THIS_KERNEL_VERSION >= LINUX(6,10,0), but no info
>> related to swap_file_is_file check?
>>
>> Thanks,
>> Tao Liu
>>
>>
>> > devname = vfsmount_devname(file_to_vfsmnt(swap_file),
>> > buf1, BUFSIZE);
>> > get_pathname(file_to_dentry(swap_file),
>> > --
>> > 2.34.1
>> >
>> > Thanks.
>> > Guanyou
>>