Re: [PATCH v2] alloc_tag: fix undetected compressed tag overflow when profiling is disabled
Hao Ge <[email protected]>
| Newsgroups | gmane.linux.kernel.stable,gmane.linux.kernel.mm,gmane.linux.kernel |
|---|---|
| Message-ID | <[email protected]> |
On 2026/8/5 03:43, Andrew Morton wrote: > On Tue, 4 Aug 2026 20:20:38 +0800 Hao Ge <[email protected]> wrote: > >> In reserve_module_tags(), the tag overflow check is gated on >> mem_alloc_profiling_enabled(): >> >> if (mem_alloc_profiling_enabled() && !tags_addressable()) >> >> If profiling is toggled off at runtime and a module is loaded whose >> tags exceed the compressed-mode limit, shutdown_mem_profiling() is >> skipped. vm_module_tags_populate() still maps memory for the tags and >> the module loads successfully, but the total tag count now exceeds what >> NR_UNUSED_PAGEFLAG_BITS can address. >> >> Once profiling is re-enabled, ref_to_idx() computes each tag's index >> as its position in the alloc_tag array. update_page_tag_ref() masks >> it to alloc_tag_ref_mask before storing in page->flags. Indices >> beyond the mask are truncated and idx_to_ref() resolves them to wrong >> tags. >> >> mem_alloc_profiling_enabled() and mem_profiling_compressed are >> independent. Once compressed mode is established at boot, it stays >> active regardless of runtime toggles of mem_profiling. >> >> Remove the mem_alloc_profiling_enabled() guard. Also return an error >> after shutdown_mem_profiling() to skip vm_module_tags_populate(), as >> the mapped pages would never be reused - shutdown_mem_profiling() sets >> mem_profiling_support to false, so no future module load enters the >> codetag path. > > Thanks. Sashiko pointed at a few things, one pre-existing. > https://sashiko.dev/#/patchset/[email protected] > > Is there any sanity in there? Hi Andrew Thanks for pointing these out. I went through all the Sashiko feedback from v1 and v2. Most of what it flagged is pre-existing and I'll fix it separately. The only point that actually touches this patch is the -ENOMEM failing the first load. Between wasting a few KB of tag memory to let the load succeed and returning -ENOMEM so the user just reloads the module once, I prefer the latter: a re-insmod works because the codetag section is skipped once profiling is off, whereas those mapped tag pages would never be used or reclaimed. Thanks Best Regards Hao