Hi, Kazu
Thank you for the comment.

On Mon, May 31, 2021 at 9:42 AM HAGIO KAZUHITO(萩尾 一仁) <k-hagio-ab@nec.com> wrote:
Hi Lianbo,

> No, the patch fixes the above errors.  Upstream kernel has changed the calculation formula in the
> freelist_ptr() as below(marked it with "^^^^"), which added the "swab()" operation, crash also needs
> to follow up this change, otherwise crash will get the error of "invalid freepointer".

> Before:
> crash> kmem -s
> CACHE             OBJSIZE  ALLOCATED     TOTAL  SLABS  SSIZE  NAME
> c0000000311a4200      152          0         0      0    64k  fuse_request
> c0000000311a8700      800          0         0      0    64k  fuse_inode
> c000000035df7600      528          0         0      0    64k  xfs_dqtrx
> c000000035df6600      496          0         0      0    64k  xfs_dquot
> kmem: xfs_buf: slab: c00c0000000b6500 invalid freepointer: 3808942d00004eb0
> kmem: xfs_buf: slab: c00c00000116ba00 invalid freepointer: b83fe85a040027b0
> ...

Yes, you are right, and thanks for the info.

I was confused because I could not reproduce the error message with a dozen
vmcores.

 
 The above error messages are easily observed on ppc64le and s390 machines.

With some debugging, I think maybe the message itself does not occur on x86_64
due to this issue, because get_freepointer() returns a value having 1 in bit 1
on x86_64 and breaks the loop below.

@@ -19361,6 +19361,7 @@ get_freepointer(struct meminfo *si, void *object)
                return BADADDR;
        }

+       fprintf(fp, "DEBUG: vaddr: %lx ptr: %lx\n", vaddr, freelist_ptr(si,nextfree,vaddr));
        return (freelist_ptr(si, nextfree, vaddr));
 }

crash> kmem -s
CACHE             OBJSIZE  ALLOCATED     TOTAL  SLABS  SSIZE  NAME
...
DEBUG: vaddr: ffff927b0501b5b8 ptr: b8b501057b92ec47
DEBUG: vaddr: ffff927b05221bb8 ptr: b81b22057b92f8c7
DEBUG: vaddr: ffff927b05218e38 ptr: 388e21057b92f647
DEBUG: vaddr: ffff927b052269b8 ptr: b86922057b92f647
                                                   ^bit1 = 1
ffff927b08452200      360        802       840     40     8k  xfs_buf
DEBUG: vaddr: ffff927b07991900 ptr: 1999077b92f1f7
DEBUG: vaddr: ffff927b017ac130 ptr: 30c17a017b92f017
DEBUG: vaddr: ffff927b0496d9c8 ptr: c8d996047b92fcc7
DEBUG: vaddr: ffff927b01bce5e0 ptr: e0e5bc017b92f37f
ffff927b08452b00      200         50       480     24     4k  xfs_bui_item

        for (q = freelist; q; q = get_freepointer(si, (void *)q)) {
                if (q & PAGE_MAPPING_ANON)
                        break;  <<-- break here


However, the issue affects the calculation of allocated/free objects
also on x86_64:

-ffff927b08452200      360        802       840     40     8k  xfs_buf
+ffff927b08452200      360        777       840     40     8k  xfs_buf
                                  ^^^

>       > > Linux 5.7 and later kernels that contain kernel commit <1ad53d9fa3f6>
>       > > ("slub: improve bit diffusion for freelist ptr obfuscation") changed
>       > > the calculation formula in the freelist_ptr(), which added a swab()
>       > > call to mix bits a little more. When kernel is built with the
>       > > "CONFIG_SLAB_FREELIST_HARDENED=y",the "kmem -s" option fails with the
>       > > following errors, if there is no such patch.

So, could I change the latter part of the commit message to the following?


Yes,  this also looks good.

Thanks.
Lianbo
   
  When kernel is configured with the "CONFIG_SLAB_FREELIST_HARDENED=y",
  Without the patch, the "kmem -s|-S" option displays wrong statistics and
  state whether the slab objects are in use or free and can print the
  following errors.

Thanks,
Kazu