| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <dcbe4f3e-42f3-d245-9ee4-b17977ad27a5@redhat.com>
Date: Thu, 5 Nov 2020 11:26:03 +0800
From: Jason Wang <jasowang@...hat.com>
To: Si-Wei Liu <si-wei.liu@...cle.com>, mst@...hat.com,
lingshan.zhu@...el.com
Cc: joao.m.martins@...cle.com, boris.ostrovsky@...cle.com,
linux-kernel@...r.kernel.org,
virtualization@...ts.linux-foundation.org
Subject: Re: [PATCH v2] vhost-vdpa: fix page pinning leakage in error path
(rework)
On 2020/11/5 上午7:33, Si-Wei Liu wrote:
> Pinned pages are not properly accounted particularly when
> mapping error occurs on IOTLB update. Clean up dangling
> pinned pages for the error path.
>
> The memory usage for bookkeeping pinned pages is reverted
> to what it was before: only one single free page is needed.
> This helps reduce the host memory demand for VM with a large
> amount of memory, or in the situation where host is running
> short of free memory.
>
> Fixes: 4c8cf31885f6 ("vhost: introduce vDPA-based backend")
> Signed-off-by: Si-Wei Liu <si-wei.liu@...cle.com>
> ---
> Changes in v2:
> - Drop the reversion patch
> - Fix unhandled page leak towards the end of page_list
>
> drivers/vhost/vdpa.c | 79 ++++++++++++++++++++++++++++++++++++++++------------
> 1 file changed, 61 insertions(+), 18 deletions(-)
>
> diff --git a/drivers/vhost/vdpa.c b/drivers/vhost/vdpa.c
> index b6d9016..e112854 100644
> --- a/drivers/vhost/vdpa.c
> +++ b/drivers/vhost/vdpa.c
> @@ -560,6 +560,8 @@ static int vhost_vdpa_map(struct vhost_vdpa *v,
>
> if (r)
> vhost_iotlb_del_range(dev->iotlb, iova, iova + size - 1);
> + else
> + atomic64_add(size >> PAGE_SHIFT, &dev->mm->pinned_vm);
>
> return r;
> }
> @@ -591,14 +593,16 @@ static int vhost_vdpa_process_iotlb_update(struct vhost_vdpa *v,
> unsigned long list_size = PAGE_SIZE / sizeof(struct page *);
> unsigned int gup_flags = FOLL_LONGTERM;
> unsigned long npages, cur_base, map_pfn, last_pfn = 0;
> - unsigned long locked, lock_limit, pinned, i;
> + unsigned long lock_limit, sz2pin, nchunks, i;
> u64 iova = msg->iova;
> + long pinned;
> int ret = 0;
>
> if (vhost_iotlb_itree_first(iotlb, msg->iova,
> msg->iova + msg->size - 1))
> return -EEXIST;
>
> + /* Limit the use of memory for bookkeeping */
> page_list = (struct page **) __get_free_page(GFP_KERNEL);
> if (!page_list)
> return -ENOMEM;
> @@ -607,52 +611,75 @@ static int vhost_vdpa_process_iotlb_update(struct vhost_vdpa *v,
> gup_flags |= FOLL_WRITE;
>
> npages = PAGE_ALIGN(msg->size + (iova & ~PAGE_MASK)) >> PAGE_SHIFT;
> - if (!npages)
> - return -EINVAL;
> + if (!npages) {
> + ret = -EINVAL;
> + goto free;
> + }
>
> mmap_read_lock(dev->mm);
>
> - locked = atomic64_add_return(npages, &dev->mm->pinned_vm);
> lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
> -
> - if (locked > lock_limit) {
> + if (npages + atomic64_read(&dev->mm->pinned_vm) > lock_limit) {
> ret = -ENOMEM;
> - goto out;
> + goto unlock;
> }
>
> cur_base = msg->uaddr & PAGE_MASK;
> iova &= PAGE_MASK;
> + nchunks = 0;
>
> while (npages) {
> - pinned = min_t(unsigned long, npages, list_size);
> - ret = pin_user_pages(cur_base, pinned,
> - gup_flags, page_list, NULL);
> - if (ret != pinned)
> + sz2pin = min_t(unsigned long, npages, list_size);
> + pinned = pin_user_pages(cur_base, sz2pin,
> + gup_flags, page_list, NULL);
> + if (sz2pin != pinned) {
> + if (pinned < 0) {
> + ret = pinned;
> + } else {
> + unpin_user_pages(page_list, pinned);
> + ret = -ENOMEM;
> + }
> goto out;
> + }
> + nchunks++;
>
> if (!last_pfn)
> map_pfn = page_to_pfn(page_list[0]);
>
> - for (i = 0; i < ret; i++) {
> + for (i = 0; i < pinned; i++) {
> unsigned long this_pfn = page_to_pfn(page_list[i]);
> u64 csize;
>
> if (last_pfn && (this_pfn != last_pfn + 1)) {
> /* Pin a contiguous chunk of memory */
> csize = (last_pfn - map_pfn + 1) << PAGE_SHIFT;
> - if (vhost_vdpa_map(v, iova, csize,
> - map_pfn << PAGE_SHIFT,
> - msg->perm))
> + ret = vhost_vdpa_map(v, iova, csize,
> + map_pfn << PAGE_SHIFT,
> + msg->perm);
> + if (ret) {
> + /*
> + * Unpin the pages that are left unmapped
> + * from this point on in the current
> + * page_list. The remaining outstanding
> + * ones which may stride across several
> + * chunks will be covered in the common
> + * error path subsequently.
> + */
> + unpin_user_pages(&page_list[i],
> + pinned - i);
Can we simply do last_pfn = this_pfn here?
> goto out;
> + }
> +
> map_pfn = this_pfn;
> iova += csize;
> + nchunks = 0;
> }
>
> last_pfn = this_pfn;
> }
>
> - cur_base += ret << PAGE_SHIFT;
> - npages -= ret;
> + cur_base += pinned << PAGE_SHIFT;
> + npages -= pinned;
> }
>
> /* Pin the rest chunk */
> @@ -660,10 +687,26 @@ static int vhost_vdpa_process_iotlb_update(struct vhost_vdpa *v,
> map_pfn << PAGE_SHIFT, msg->perm);
> out:
> if (ret) {
> + if (nchunks && last_pfn) {
Any reason for checking last_pfn here?
Note that we did:
+ nchunks++;
if (!last_pfn)
map_pfn = page_to_pfn(page_list[0]);
Thanks
> + unsigned long pfn;
> +
> + /*
> + * Unpin the outstanding pages which are yet to be
> + * mapped but haven't due to vdpa_map() or
> + * pin_user_pages() failure.
> + *
> + * Mapped pages are accounted in vdpa_map(), hence
> + * the corresponding unpinning will be handled by
> + * vdpa_unmap().
> + */
> + for (pfn = map_pfn; pfn <= last_pfn; pfn++)
> + unpin_user_page(pfn_to_page(pfn));
> + }
> vhost_vdpa_unmap(v, msg->iova, msg->size);
> - atomic64_sub(npages, &dev->mm->pinned_vm);
> }
> +unlock:
> mmap_read_unlock(dev->mm);
> +free:
> free_page((unsigned long)page_list);
> return ret;
> }
Powered by blists - more mailing lists