[PATCH v2 0/3] EAL memory fixes

[Jake Freeland]

On Thu Jan 22, 2026 at 11:09 PM CST, Stephen Hemminger wrote:
> On Thu, 14 Aug 2025 16:32:42 -0500
> Jake Freeland <jfree at FreeBSD.org> wrote:
>
>> Hi there,
>> 
>> This patchset contains a number of EAL-specific memory fixes that I've
>> made over the last year. Two pertain to FreeBSD, one pertains to Linux.
>> 
>> v2:
>> * Log messages are no longer split across multiple lines.
>> * The patch titled "Do not index out of bounds in memseg list" has been
>>   removed since there was no indexing out of bounds happening.
>> * A new patch has been added per Anatoly's recommendation that starts
>>   searching for memseg entries after the last used entry.
>> 
>> Jake Freeland (3):
>>   eal/freebsd: Do not use prev_ms_idx for hole detection
>>   eal/freebsd: Avoid claiming memseg holes
>>   eal/linux: Check hugepage access permissions
>> 
>>  lib/eal/freebsd/eal_memory.c      | 24 +++++++++++++++++++-----
>>  lib/eal/linux/eal_hugepage_info.c |  7 +++++++
>>  2 files changed, 26 insertions(+), 5 deletions(-)
>> 
>> --
>> 2.47.2
>> 
>
> AI review of this series is mostly positive
>
> === Patch Review: bundle-1679-eal-freebsd.mbox (via Claude) ===
> # DPDK Patch Review: eal/freebsd Memory Management Fixes
>
> ## Summary
> This patch series contains 3 commits addressing memory management issues in FreeBSD EAL and adding permission checks for Linux hugepage directories.
>
> ---
>
> ## Patch 1/3: eal/freebsd: Do not use prev_ms_idx for hole detection
>
> ### Errors
> None found.
>
> ### Warnings
>
> 1. **Subject line capitalization issue**
>    - Subject: `eal/freebsd: Do not use prev_ms_idx for hole detection`
>    - "Do" should be lowercase after the colon per guidelines
>    - Should be: `eal/freebsd: do not use prev_ms_idx for hole detection`
>
> ### Info
> - Commit message is clear and explains the problem well
> - `Signed-off-by` and `Acked-by` tags present and in correct order
> - Code change is minimal and focused
> - Line lengths are within limits
>
> ---
>
> ## Patch 2/3: eal/freebsd: Avoid claiming memseg holes
>
> ### Errors
> None found.
>
> ### Warnings
>
> 1. **Subject line capitalization issue**
>    - Subject: `eal/freebsd: Avoid claiming memseg holes`
>    - "Avoid" should be lowercase after the colon
>    - Should be: `eal/freebsd: avoid claiming memseg holes`
>
> 2. **Variable declaration style inconsistency**
>    ```c
>    for (msl_idx = 0; msl_idx < RTE_MAX_MEMSEG_LISTS;
>    				msl_idx++) {
>    	int start_idx, num_elems;
>    	bool empty, need_hole;
>    ```
>    - The new variables `start_idx` and `num_elems` are declared at the start of the block, which is acceptable. However, they could be declared closer to their point of use or initialized when meaningful values exist.
>
> 3. **Potential logic issue with start_idx check**
>    ```c
>    start_idx = rte_fbarray_find_prev_n_used(
>    		arr, arr->len - 1, 1) + 1;
>    if (start_idx < 0)
>    	start_idx = 0;
>    ```
>    - Since `rte_fbarray_find_prev_n_used()` returns -1 on failure, and then +1 is added, `start_idx` would be 0 when the function fails, making the `if (start_idx < 0)` check unreachable. This appears to be dead code. Consider:
>    ```c
>    start_idx = rte_fbarray_find_prev_n_used(arr, arr->len - 1, 1);
>    if (start_idx < 0)
>    	start_idx = 0;
>    else
>    	start_idx++;
>    ```
>
> ### Info
> - Commit message explains the fix adequately
> - Tags are present and in correct order
>
> ---
>
> ## Patch 3/3: eal/linux: Check hugepage access permissions
>
> ### Errors
> None found.
>
> ### Warnings
>
> 1. **Subject line capitalization issue**
>    - Subject: `eal/linux: Check hugepage access permissions`
>    - "Check" should be lowercase after the colon
>    - Should be: `eal/linux: check hugepage access permissions`
>
> 2. **Use of lowercase log message**
>    - The log message uses lowercase which is inconsistent with some other EAL log messages, though this is a minor style preference.
>
> ### Info
>
> 1. **Consider using `faccessat()` for race-free checks**
>    - The `access()` function has TOCTOU (time-of-check-time-of-use) race conditions. While this is likely acceptable for this use case (startup configuration), it's worth noting.
>
> 2. **Good defensive programming**
>    - Adding permission checks before attempting to use mountpoints improves error handling and user feedback.
>
> 3. **Commit message is clear and concise**
>
> ---
>
> ## General Observations
>
> ### All Patches
> - All patches have proper `Signed-off-by` tags with real names and valid email addresses
> - All patches have `Acked-by` from the maintainer
> - Tag order is correct (Signed-off-by before Acked-by)
> - No trailing whitespace detected
> - Line lengths appear to be within 100 character limit for code
> - Commit body lines are within 75 character limit
> - Files have appropriate licenses (existing files being modified)
>
> ### Missing Items (Info level)
> - No `Fixes:` tag on patches 1 and 2, which appear to be bug fixes. If these fix regressions from specific commits, a `Fixes:` tag should be added.
> - No `Cc: stable at dpdk.org` tag if these are candidates for backporting to stable releases.
>
> ---
>
> ## Summary of Required Changes
>
> | Severity | Count | Items |
> |----------|-------|-------|
> | Error | 0 | - |
> | Warning | 4 | 3x subject line capitalization, 1x potential dead code |
> | Info | 4 | Missing Fixes tags, TOCTOU note, style preferences |
>
> ### Recommended Actions
>
> 1. **Fix subject lines** (all 3 patches): Change to lowercase after colon
>    - `eal/freebsd: do not use prev_ms_idx for hole detection`
>    - `eal/freebsd: avoid claiming memseg holes`
>    - `eal/linux: check hugepage access permissions`
>
> 2. **Review the start_idx logic** in patch 2 to ensure the `< 0` check is reachable
>
> 3. **Consider adding `Fixes:` tags** if these address specific regressions

Hi Stephen,

This AI analysis seems to have been run on the previous revision on the
patchset.

Jake Freeland