Re: [PATCH v2 4/5] proc/task_mmu: read proc/pid/smaps_rollup under per-vma lock
From: Suren Baghdasaryan
Date: Wed Sep 09 2026 - 13:37:44 EST
On Wed, Sep 9, 2026 at 7:25 AM Usama Arif <usama.arif@xxxxxxxxx> wrote:
>
> On Sun, 6 Sep 2026 23:39:17 -0700 Suren Baghdasaryan <surenb@xxxxxxxxxx> wrote:
>
> > proc/pid/smaps_rollup can be read using the combination of RCU and
> > VMA read locks, similar to proc/pid/{maps|smaps|numa_maps}. RCU is
> > required to safely traverse the VMA tree and VMA lock stabilizes the
> > VMA being processed and the pagetable walk.
> > Note that we have to keep the logic to drop mmap_lock on contention
> > because even when using per-VMA locks we might have to fall back to
> > holding the mmap_lock.
> >
> > Running Paul's contention benchmark [1] shows considerable improvement
> > both in median and in the worst case latencies:
> >
> > Execution command: run-proc-vs-map.sh --nsamples 20 --rawdata -- \
> > --busyduration 2 --procfile smaps_rollup
> >
> > Baseline:
> > Median Minimum Maximum
> > 0.174 0.161 2.553
> > 0.174 0.164 2.663
> > 0.174 0.165 2.664
> > 0.174 0.166 2.679
> > 0.174 0.167 2.691
> > 0.174 0.168 2.704
> > 0.174 0.169 2.729
> > 0.174 0.172 2.741
> > 0.174 0.174 2.745
> > 0.174 0.174 2.755
> > 0.174 0.175 2.790
> > 0.174 0.177 2.809
> > 0.174 0.179 3.096
> > 0.174 0.183 3.144
> > 0.174 0.184 3.158
> > 0.174 0.185 3.175
> > 0.174 0.185 4.568
> > 0.174 0.198 4.821
> > 0.174 0.214 5.143
> > 0.174 0.251 5.220
> >
> > Patched:
> > Median Minimum Maximum
> > 0.007 0.007 1.952
> > 0.007 0.007 1.955
> > 0.007 0.007 1.955
> > 0.007 0.007 1.955
> > 0.007 0.007 1.957
> > 0.007 0.007 1.969
> > 0.007 0.007 2.065
> > 0.007 0.007 2.075
> > 0.007 0.007 2.146
> > 0.007 0.007 2.195
> > 0.007 0.007 2.223
> > 0.007 0.007 2.259
> > 0.007 0.007 2.488
> > 0.007 0.007 2.562
> > 0.007 0.007 2.599
> > 0.007 0.007 2.697
> > 0.007 0.007 3.030
> > 0.007 0.007 3.075
> > 0.007 0.007 3.145
> > 0.007 0.007 3.225
> >
> > Remove now unused lock_ctx_mm() and move unlock_ctx_vma() next to
> > unlock_ctx_mm() as they are logically related.
> >
> > Remove a long comment about 4 cases that we handle when dropping the
> > mmap lock in the middle of VMA walk due to contention. The first 3
> > cases explained there are handled naturally and only case 4 needs to
> > be handled in a special way, which is done in smap_gather_stats() by
> > gathering stats from the portion of the VMA that has not yet been
> > processed.
> > For posterity, moving this comment here:
> >
> > After dropping the lock, there are four cases to
> > consider. See the following example for explanation.
> >
> > +------+------+-----------+
> > | VMA1 | VMA2 | VMA3 |
> > +------+------+-----------+
> > | | | |
> > 4k 8k 16k 400k
> >
> > Suppose we drop the lock after reading VMA2 due to
> > contention, then we get:
> >
> > last_vma_end = 16k
> >
> > 1) VMA2 is freed, but VMA3 exists:
> >
> > vma_next(vmi) will return VMA3.
> > In this case, just continue from VMA3.
> >
> > 2) VMA2 still exists:
> >
> > vma_next(vmi) will return VMA3.
> > In this case, just continue from VMA3.
> >
> > 3) No more VMAs can be found:
> >
> > vma_next(vmi) will return NULL.
> > No more things to do, just break.
> >
> > 4) (last_vma_end - 1) is the middle of a vma (VMA'):
> >
> > vma_next(vmi) will return VMA' whose range
> > contains last_vma_end.
> > Iterate VMA' from last_vma_end.
> >
> > [1] https://github.com/paulmckrcu/proc-mmap_sem-test
> >
> > Signed-off-by: Suren Baghdasaryan <surenb@xxxxxxxxxx>
> > ---
> > fs/proc/task_mmu.c | 153 ++++++++++++++++++---------------------------
> > 1 file changed, 60 insertions(+), 93 deletions(-)
> >
> > diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
> > index 3351decd1172..641a155b0c61 100644
> > --- a/fs/proc/task_mmu.c
> > +++ b/fs/proc/task_mmu.c
> > @@ -130,28 +130,12 @@ static void release_task_mempolicy(struct proc_maps_private *priv)
> > }
> > #endif
> >
> > -static int lock_ctx_mm(struct proc_maps_locking_ctx *lock_ctx)
> > -{
> > - int ret = mmap_read_lock_killable(lock_ctx->mm);
> > -
> > - if (!ret)
> > - lock_ctx->mmap_locked = true;
> > -
> > - return ret;
> > -}
> > -
> > static void unlock_ctx_mm(struct proc_maps_locking_ctx *lock_ctx)
> > {
> > mmap_read_unlock(lock_ctx->mm);
> > lock_ctx->mmap_locked = false;
> > }
> >
> > -static void reset_lock_ctx(struct proc_maps_locking_ctx *lock_ctx)
> > -{
> > - lock_ctx->locked_vma = NULL;
> > - lock_ctx->mmap_locked = false;
> > -}
> > -
> > static void unlock_ctx_vma(struct proc_maps_locking_ctx *lock_ctx)
> > {
> > if (lock_ctx->locked_vma) {
> > @@ -160,6 +144,12 @@ static void unlock_ctx_vma(struct proc_maps_locking_ctx *lock_ctx)
> > }
> > }
> >
> > +static void reset_lock_ctx(struct proc_maps_locking_ctx *lock_ctx)
> > +{
> > + lock_ctx->locked_vma = NULL;
> > + lock_ctx->mmap_locked = false;
> > +}
> > +
> > static struct vm_area_struct *get_next_vma(struct proc_maps_private *priv,
> > loff_t last_pos)
> > {
> > @@ -1376,12 +1366,14 @@ static int show_smap(struct seq_file *m, void *v)
> > static int show_smaps_rollup(struct seq_file *m, void *v)
> > {
> > struct proc_maps_private *priv = m->private;
> > + struct proc_maps_locking_ctx *lock_ctx = &priv->lock_ctx;
> > + struct mm_struct *mm = lock_ctx->mm;
> > struct mem_size_stats mss = {};
> > - struct mm_struct *mm = priv->lock_ctx.mm;
> > + unsigned long last_vma_end = 0;
> > + unsigned long vma_start = 0;
> > struct vm_area_struct *vma;
> > - unsigned long vma_start = 0, last_vma_end = 0;
> > + loff_t pos = 0;
> > int ret = 0;
> > - VMA_ITERATOR(vmi, mm, 0);
> >
> > priv->task = get_proc_task(priv->inode);
> > if (!priv->task)
> > @@ -1392,89 +1384,60 @@ static int show_smaps_rollup(struct seq_file *m, void *v)
> > goto out_put_task;
> > }
> >
> > - ret = lock_ctx_mm(&priv->lock_ctx);
> > - if (ret)
> > - goto out_put_mm;
> > -
> > hold_task_mempolicy(priv);
> > - vma = vma_next(&vmi);
> > + rcu_read_lock();
> > + reset_lock_ctx(lock_ctx);
> >
> > + vma_iter_init(&priv->iter, mm, 0);
> > + vma = proc_get_vma(m, &pos);
> > if (unlikely(!vma))
> > goto empty_set;
> >
> > - vma_start = vma->vm_start;
> > - do {
> > - smap_gather_stats(priv, vma, &mss, vma->vm_start);
> > - last_vma_end = vma->vm_end;
> > + if (!IS_ERR(vma) && vma != get_gate_vma(lock_ctx->mm))
> > + vma_start = vma->vm_start;
> > +
> > + while (vma) {
> > + if (IS_ERR(vma)) {
> > + ret = PTR_ERR(vma);
> > + goto out_unlock;
> > + }
> > +
> > + if (vma == get_gate_vma(lock_ctx->mm))
> > + break;
> >
> > /*
> > - * Release mmap_lock temporarily if someone wants to
> > - * access it for write request.
> > + * If after retaking the lock, already reported VMA grew or
> > + * merged with the next one, smap_gather_stats() will gather
> > + * stats for the remaining portion by starting at last_vma_end.
> > */
> > - if (mmap_lock_is_contended(mm)) {
> > - vma_iter_invalidate(&vmi);
> > - unlock_ctx_mm(&priv->lock_ctx);
> > - ret = lock_ctx_mm(&priv->lock_ctx);
> > - if (ret) {
> > - release_task_mempolicy(priv);
> > - goto out_put_mm;
> > - }
> > + smap_gather_stats(priv, vma, &mss, last_vma_end);
>
> Patch 3 made smap_gather_stats() reject starts below the VMA, while this
> function initializes last_vma_end to zero.
>
> The first ordinary VMA is therefore skipped. lock_next_vma() can also
> return a VMA beginning after the requested position, so the first VMA
> after every unmapped gap is skipped as well. This causes
> smaps_rollup to underreport RSS, PSS, swap, and the other accumulated
> values.
>
> I think you need:
> unsigned long start = max(last_vma_end, vma->vm_start);
> smap_gather_stats(priv, vma, &mss, start);
Oops, you are right. I didn't realize the first VMA would be silently
skipped. I'll fix it. Thanks!
>
>
>
> > + last_vma_end = vma->vm_end;
> >
> > + /*
> > + * If the VMA lock is not taken, we hold the often contended
> > + * mmap lock. This can happen if we had to fall back to the
> > + * mmap lock.
> > + *
> > + * To relieve pressure, check if it is indeed contended, then
> > + * temporarily release it.
> > + */
> > + if (lock_ctx->mmap_locked &&
> > + mmap_lock_is_contended(lock_ctx->mm)) {
> > + unlock_ctx_mm(lock_ctx);
> > /*
> > - * After dropping the lock, there are four cases to
> > - * consider. See the following example for explanation.
> > - *
> > - * +------+------+-----------+
> > - * | VMA1 | VMA2 | VMA3 |
> > - * +------+------+-----------+
> > - * | | | |
> > - * 4k 8k 16k 400k
> > - *
> > - * Suppose we drop the lock after reading VMA2 due to
> > - * contention, then we get:
> > - *
> > - * last_vma_end = 16k
> > - *
> > - * 1) VMA2 is freed, but VMA3 exists:
> > - *
> > - * vma_next(vmi) will return VMA3.
> > - * In this case, just continue from VMA3.
> > - *
> > - * 2) VMA2 still exists:
> > - *
> > - * vma_next(vmi) will return VMA3.
> > - * In this case, just continue from VMA3.
> > - *
> > - * 3) No more VMAs can be found:
> > - *
> > - * vma_next(vmi) will return NULL.
> > - * No more things to do, just break.
> > - *
> > - * 4) (last_vma_end - 1) is the middle of a vma (VMA'):
> > - *
> > - * vma_next(vmi) will return VMA' whose range
> > - * contains last_vma_end.
> > - * Iterate VMA' from last_vma_end.
> > + * Even though we previously fell back to mmap lock,
> > + * we try taking VMA lock for the next VMA, since it
> > + * might not be under modification. In the worst case
> > + * we will fall back to mmap lock again.
> > */
> > - vma = vma_next(&vmi);
> > - /* Case 3 above */
> > - if (!vma)
> > - break;
> > -
> > - /* Case 1 and 2 above */
> > - if (vma->vm_start >= last_vma_end) {
> > - smap_gather_stats(priv, vma, &mss, vma->vm_start);
> > - last_vma_end = vma->vm_end;
> > - continue;
> > - }
> > -
> > - /* Case 4 above */
> > - if (vma->vm_end > last_vma_end) {
> > - smap_gather_stats(priv, vma, &mss, last_vma_end);
> > - last_vma_end = vma->vm_end;
> > - }
> > + rcu_read_lock();
> > + reset_lock_ctx(lock_ctx);
> > + /* Resume from the last position. */
> > + pos = last_vma_end;
> > + vma_iter_init(&priv->iter, mm, pos);
> > }
> > - } for_each_vma(vmi, vma);
> > + vma = proc_get_vma(m, &pos);
> > + }
> >
> > empty_set:
> > show_vma_header_prefix(m, vma_start, last_vma_end, 0, 0, 0, 0);
> > @@ -1483,10 +1446,14 @@ static int show_smaps_rollup(struct seq_file *m, void *v)
> >
> > __show_smap(m, &mss, true);
> >
> > +out_unlock:
> > + if (lock_ctx->mmap_locked) {
> > + unlock_ctx_mm(lock_ctx);
> > + } else {
> > + unlock_ctx_vma(lock_ctx);
> > + rcu_read_unlock();
> > + }
> > release_task_mempolicy(priv);
> > - unlock_ctx_mm(&priv->lock_ctx);
> > -
> > -out_put_mm:
> > mmput(mm);
> > out_put_task:
> > put_task_struct(priv->task);
> > --
> > 2.55.0.979.g7e5102b832-goog
> >
> >