run_vmtests.sh can reach test setup after invalid options or category
selections. Its memfd_secret preparation can also change ptrace_scope
when that category was not selected.

Patch 1 rejects invalid selections before setup, using the category list
in usage(). Patch 2 gates memfd_secret preparation on category selection
and executable presence.

I am resending from my Gmail address after delivery problems with my
previous address. I could not confirm delivery of the earlier v3 posting.
This series is rebased onto current mm-unstable; both patches are unchanged
from v3 according to range-diff. Neither fix is present in current
mm-unstable, mm-stable or mainline.

Changes in v3:
- Initialize VM_SELFTEST_ITEMS to "default" before getopts, as David
  suggested. Only -t changes the selection.
- Use Assisted-by: LLM in both patches.
- Add David's Acked-by to patch 2, whose code is unchanged from v2.

Validation after rebasing:
- Bash syntax, whitespace and checkpatch checks passed.
- Ten invalid option/category invocations and help passed before setup.
- Rebuilt the four focused test binaries and ran the rebased script in an
  x86-64 KVM guest. mmap, memfd_secret, combined selections, tab/newline
  separators and repeated -t with the final selection taking effect all
  passed.
- The guest lacks Yama. The ptrace_scope write condition was checked with
  a disposable file at the expected path, not Yama policy enforcement.

This is focused validation, not a run of the complete MM selftest suite.

v2: https://lore.kernel.org/r/178886112560.138404.1278745290043112298.mm-cli-v2-0@tychen.cc

Tianyi Chen (2):
  selftests/mm: Reject invalid test selections before running tests
  selftests/mm: Only prepare ptrace_scope when memfd_secret is selected

Hi David,

Following up on your review:
https://lore.kernel.org/r/d6ca7f47-39bb-43b4-92d6-63ccc148ebdd@kernel.org

> Who would set it at this point?
> Can't this just be
> VM_SELFTEST_ITEMS="default"

V3 makes that change: the default is initialized before getopts, and only
-t changes the selection. An explicit empty -t still fails validation.
Both patches use Assisted-by: LLM, and patch 2 carries your Acked-by.

I have resent v3 at the start of this new thread from my Gmail address
after delivery problems with my previous address. The patches are rebased
onto current mm-unstable without changes, and the focused parser and VM
selection tests passed again.

Thanks for the review and the ack.

Tianyi

On 9/9/26 16:54, Tianyi Chen wrote:
> Hi David,
> 
> Following up on your review:
> https://lore.kernel.org/r/d6ca7f47-39bb-43b4-92d6-63ccc148ebdd@kernel.org
> 
>> Who would set it at this point?
>> Can't this just be
>> VM_SELFTEST_ITEMS="default"
> 
> V3 makes that change: the default is initialized before getopts, and only
> -t changes the selection. An explicit empty -t still fails validation.
> Both patches use Assisted-by: LLM, and patch 2 carries your Acked-by.
> 
> I have resent v3 at the start of this new thread from my Gmail address
> after delivery problems with my previous address. The patches are rebased
> onto current mm-unstable without changes, and the focused parser and VM
> selection tests passed again.

That is new:

$ b4 shazam https://lore.kernel.org/r/CACGbirR9PNt4uhqJFATsOWnwmsNb-v8zau-Rb+9Q5SZmY-=Vug@mail.gmail.com
Looking up https://lore.kernel.org/all/CACGbirR9PNt4uhqJFATsOWnwmsNb-v8zau-Rb%2B9Q5SZmY-%3DVug@mail.gmail.com/
Grabbing thread from lore.kernel.org/all/CACGbirR9PNt4uhqJFATsOWnwmsNb-v8zau-Rb%2B9Q5SZmY-%3DVug@mail.gmail.com/t.mbox.gz
Checking for newer revisions
Grabbing search results from lore.kernel.org
Analyzing 4 messages in the thread
Analyzing 14 code-review messages
Checking attestation on all messages, may take a moment...
---
  ✓ [PATCH v3 1/2] selftests/mm: Reject invalid test selections before running tests
  ✓ [PATCH v3 2/2] selftests/mm: Only prepare ptrace_scope when memfd_secret is selected
  ---
  ✓ Signed: DKIM/gmail.com
---
Total patches: 2
---
Applying: selftests/mm: Reject invalid test selections before running tests
Patch failed at 0001 selftests/mm: Reject invalid test selections before running tests
error: git diff header lacks filename information when removing 1 leading pathname component at /home/dishy/linux/.git/rebase-apply/patch:6


-- 
Cheers,

David

On 9/9/26 17:02, David Hildenbrand (Arm) wrote:
> On 9/9/26 16:54, Tianyi Chen wrote:
>> Hi David,
>>
>> Following up on your review:
>> https://lore.kernel.org/r/d6ca7f47-39bb-43b4-92d6-63ccc148ebdd@kernel.org
>>
>>> Who would set it at this point?
>>> Can't this just be
>>> VM_SELFTEST_ITEMS="default"
>>
>> V3 makes that change: the default is initialized before getopts, and only
>> -t changes the selection. An explicit empty -t still fails validation.
>> Both patches use Assisted-by: LLM, and patch 2 carries your Acked-by.
>>
>> I have resent v3 at the start of this new thread from my Gmail address
>> after delivery problems with my previous address. The patches are rebased
>> onto current mm-unstable without changes, and the focused parser and VM
>> selection tests passed again.
> 
> That is new:
> 
> $ b4 shazam https://lore.kernel.org/r/CACGbirR9PNt4uhqJFATsOWnwmsNb-v8zau-Rb+9Q5SZmY-=Vug@mail.gmail.com
> Looking up https://lore.kernel.org/all/CACGbirR9PNt4uhqJFATsOWnwmsNb-v8zau-Rb%2B9Q5SZmY-%3DVug@mail.gmail.com/
> Grabbing thread from lore.kernel.org/all/CACGbirR9PNt4uhqJFATsOWnwmsNb-v8zau-Rb%2B9Q5SZmY-%3DVug@mail.gmail.com/t.mbox.gz
> Checking for newer revisions
> Grabbing search results from lore.kernel.org
> Analyzing 4 messages in the thread
> Analyzing 14 code-review messages
> Checking attestation on all messages, may take a moment...
> ---
>   ✓ [PATCH v3 1/2] selftests/mm: Reject invalid test selections before running tests
>   ✓ [PATCH v3 2/2] selftests/mm: Only prepare ptrace_scope when memfd_secret is selected
>   ---
>   ✓ Signed: DKIM/gmail.com
> ---
> Total patches: 2
> ---
> Applying: selftests/mm: Reject invalid test selections before running tests
> Patch failed at 0001 selftests/mm: Reject invalid test selections before running tests
> error: git diff header lacks filename information when removing 1 leading pathname component at /home/dishy/linux/.git/rebase-apply/patch:6
> 
> 

I think what's happening is that something in your mail sending path wraps long
lines and breaks the patch format.

-- 
Cheers,

David

Hi David,

> I think what's happening is that something in your mail sending path wraps long
> lines and breaks the patch format.

You were right. I reproduced the same b4 error using an externally
received copy: outbound delivery wrapped the diff header and removed
whitespace. The copy in my Gmail Sent folder was intact, so my earlier
check missed the corruption. Sorry for the trouble.

I have resent the unchanged series as text/plain attachments:
https://lore.kernel.org/r/CACGbirTvSoSeNSpMCBiifTdRKKDsKbgk7nfvh=09fHw2b0M4Ew@mail.gmail.com

I verified the externally received attachments through b4 and git am.
They reproduce the prepared tree and retain the original authorship
and your Acked-by on patch 2.

I will use this Gmail address for follow-up, given the delivery problems
with my previous address.

Thanks,
Tianyi