Re: [PATCH 07/22] slab: Remove folio references from __ksize()

[Matthew Wilcox <willy@infradead.org>]

On Fri, Aug 29, 2025 at 04:47:11PM +0100, Matthew Wilcox (Oracle) wrote:
> This deprecates calling ksize() on memory allocated by alloc_pages().
> Today it becomes a warning and support will be removed entirely in
> the future.

We can't do this.  At least not without fixing x509:

https://lore.kernel.org/all/202509022248.dbe9cc08-lkp@intel.com/

The important bit:

[ 33.279697][ T1] WARNING: mm/slab_common.c:1016 at __ksize+0xf2/0x120, CPU#0: swapper/0/1
[ 33.341827][ T1] kfree_sensitive (mm/slab_common.c:1234)
[ 33.343783][ T1] mpi_free (lib/crypto/mpi/mpiutil.c:?)
[ 33.345494][ T1] rsa_enc (crypto/rsa.c:62)
[ 33.347230][ T1] rsassa_pkcs1_verify (include/linux/crypto.h:383 crypto/rsassa-pkcs1.c:256)
[ 33.349482][ T1] public_key_verify_signature (crypto/asymmetric_keys/public_key.c:432)
[ 33.352192][ T1] ? __kmalloc_noprof (include/linux/kernel.h:?)
[ 33.354454][ T1] ? asymmetric_key_generate_id (crypto/asymmetric_keys/asymmetric_type.c:147)
[ 33.357247][ T1] x509_check_for_self_signed (crypto/asymmetric_keys/x509_public_key.c:126)
[ 33.359744][ T1] x509_cert_parse (crypto/asymmetric_keys/x509_cert_parser.c:130)
[ 33.361690][ T1] x509_key_preparse (crypto/asymmetric_keys/x509_public_key.c:157)
[ 33.368645][ T1] asymmetric_key_preparse (crypto/asymmetric_keys/asymmetric_type.c:409)
[ 33.371160][ T1] __key_create_or_update (security/keys/key.c:858)
[ 33.373639][ T1] key_create_or_update (security/keys/key.c:1021)
[ 33.376003][ T1] x509_load_certificate_list (crypto/asymmetric_keys/x509_loader.c:31)
[ 33.378485][ T1] ? load_module_cert (certs/system_keyring.c:280)
[ 33.380708][ T1] do_one_initcall (init/main.c:1281)

so mpi_free() calls kfree_sensitive() on memory that didn't come from
kmalloc().  I don't have the stomach to go crawling through the
crypto code to figure out where it did come from, so I'm just going to
assume it was get_free_pages(), take out the warning; make ksize()
work for raw pages and document that this is supported behaviour.