[lkp] [+1158 bytes kernel size regression] [i386-tinyconfig] [22dd199a72] mm: introduce secretmemfd system call to create "secret" memory areas

[lkp] [+1158 bytes kernel size regression] [i386-tinyconfig] [22dd199a72] mm: introduce secretmemfd system call to create "secret" memory areas

[kernel test robot]

[-- Attachment #1: Type: text/plain, Size: 1816 bytes --]


FYI, we noticed a +1158 bytes kernel size regression due to commit:

commit: 22dd199a7221932b194badb999181ca3fa2e62e2 (mm: introduce secretmemfd system call to create "secret" memory areas)
https://git.kernel.org/pub/scm/linux/kernel/git/rppt/linux.git secretmemfd/v0

Details as below (size data is obtained by `nm --size-sort vmlinux`):

f56cf703: mmap: make mlock_future_check() global
22dd199a: mm: introduce secretmemfd system call to create "secret" memory areas

+--------------------------------+----------+----------+-------+
|             symbol             | f56cf703 | 22dd199a | delta |
+--------------------------------+----------+----------+-------+
| bzImage                        | 442560   | 443008   | 448   |
| nm.T.__ia32_sys_secretmemfd    | 0        | 285      | 285   |
| nm.t.secretmem_fault           | 0        | 254      | 254   |
| nm.R.secretmem_fops            | 0        | 128      | 128   |
| nm.t.secretmem_mmap            | 0        | 110      | 110   |
| nm.r.secretmem_super_ops       | 0        | 92       | 92    |
| nm.r.secretmem_aops            | 0        | 88       | 88    |
| nm.r.secretmem_vm_ops          | 0        | 52       | 52    |
| nm.t.secretmem_evict_inode     | 0        | 37       | 37    |
| nm.d.secretmem_fs              | 0        | 36       | 36    |
| nm.t.secretmem_init            | 0        | 29       | 29    |
| nm.t.secretmem_init_fs_context | 0        | 29       | 29    |
| nm.t.secretmem_migratepage     | 0        | 6        | 6     |
| nm.t.secretmem_freepage        | 0        | 5        | 5     |
| nm.b.secretmem_mnt             | 0        | 4        | 4     |
| nm.t.secretmem_isolate_page    | 0        | 3        | 3     |
+--------------------------------+----------+----------+-------+



Thanks

Re: [lkp] [+1158 bytes kernel size regression] [i386-tinyconfig] [22dd199a72] mm: introduce secretmemfd system call to create "secret" memory areas

[Josh Triplett]

[-- Attachment #1: Type: text/plain, Size: 329 bytes --]

On Thu, Jul 16, 2020 at 02:06:18AM +0800, kernel test robot wrote:
> commit: 22dd199a7221932b194badb999181ca3fa2e62e2 (mm: introduce secretmemfd system call to create "secret" memory areas)

Unrelated to the size, I'm wondering: why have this as a separate
syscall, rather than just a flag to memfd_create?

- Josh Triplett

Re: [lkp] [+1158 bytes kernel size regression] [i386-tinyconfig] [22dd199a72] mm: introduce secretmemfd system call to create "secret" memory areas

[Josh Triplett]

[-- Attachment #1: Type: text/plain, Size: 1246 bytes --]

On Thu, Jul 16, 2020 at 10:32:59AM +0300, Mike Rapoport wrote:
> Hi,
> 
> On Wed, Jul 15, 2020 at 01:32:02PM -0700, Josh Triplett wrote:
> > On Thu, Jul 16, 2020 at 02:06:18AM +0800, kernel test robot wrote:
> > > commit: 22dd199a7221932b194badb999181ca3fa2e62e2 (mm: introduce secretmemfd system call to create "secret" memory areas)
> 
> > Unrelated to the size, I'm wondering: why have this as a separate
> > syscall, rather than just a flag to memfd_create?
> 
> Kirill asked the same question but the other way around:
> 
> https://lore.kernel.org/linux-mm/20200713105812.dnwtdhsuyj3xbh4f(a)box/
> 
> I feel we are going to bikesched about it for a while :)
> 
> More seruously, the commonality with memfd_create end at the fact that
> this a file descriptor for memory.
> 
> Having this as separate syscall will be less confusing to the users, and
> I beleive, more flexible for future changes.

Fair enough. I'm not going to express a further opinion on the bikeshed
paint color.

That said, please by all means make sure that "make tinyconfig" doesn't
get any bigger with this patch (which *should* be the case as long as
it's a Kconfig option, and no new code gets added outside the Kconfig
option).

- Josh