From: cgxu <cgxu519@mykernel.net>
To: Amir Goldstein <amir73il@gmail.com>
Cc: overlayfs <linux-unionfs@vger.kernel.org>,
Linux MM <linux-mm@kvack.org>, Miklos Szeredi <miklos@szeredi.hu>,
Andrew Morton <akpm@linux-foundation.org>,
Ritesh Harjani <riteshh@linux.ibm.com>
Subject: Re: [RFC PATCH 3/3] ovl: implement stacked mmap for shared map
Date: Mon, 31 Aug 2020 21:47:07 +0800 [thread overview]
Message-ID: <e1e2c8f0-a3b8-0a3d-3093-6188b1a829f0@mykernel.net> (raw)
In-Reply-To: <CAOQ4uxisdtoccDoQe_fYUA-jXTfy0yk=gNcMSrmbkCYaeOEPuQ@mail.gmail.com>
On 8/30/20 7:33 PM, Amir Goldstein wrote:
> On Sat, Aug 29, 2020 at 12:51 PM Chengguang Xu <cgxu519@mykernel.net> wrote:
>>
>> Implement stacked mmap for shared map to keep data
>> consistency.
>>
>> Signed-off-by: Chengguang Xu <cgxu519@mykernel.net>
>> ---
>> fs/overlayfs/file.c | 120 +++++++++++++++++++++++++++++++++++++++++---
>> 1 file changed, 114 insertions(+), 6 deletions(-)
>>
>> diff --git a/fs/overlayfs/file.c b/fs/overlayfs/file.c
>> index 14ab5344a918..db5ab200d984 100644
>> --- a/fs/overlayfs/file.c
>> +++ b/fs/overlayfs/file.c
>> @@ -21,9 +21,17 @@ struct ovl_aio_req {
>> struct fd fd;
>> };
>>
>> +static vm_fault_t ovl_fault(struct vm_fault *vmf);
>> +static vm_fault_t ovl_page_mkwrite(struct vm_fault *vmf);
>> +
>> +static const struct vm_operations_struct ovl_vm_ops = {
>> + .fault = ovl_fault,
>> + .page_mkwrite = ovl_page_mkwrite,
>> +};
>> +
>
> Interesting direction, not sure if this is workable.
> I don't know enough about mm to say.
>
> But what about the rest of the operations?
> Did you go over them and decide that overlay doesn't need to implement them?
> I doubt it, but if you did, please document that.
I did some check for rest of them, IIUC ->fault will be enough for this
special case (shared read-only mmap with no upper), I will remove
->page_mkwrite in v2.
# I do not consider support ->huge_fault in current stage due to many fs
cannot support DAX properly.
BTW, do you know who should I add to CC list for further deep review of
this code? fadevel-list?
>
>> struct ovl_file_entry {
>> struct file *realfile;
>> - void *vm_ops;
>> + const struct vm_operations_struct *vm_ops;
>> };
>>
>> struct file *ovl_get_realfile(struct file *file)
>> @@ -40,14 +48,15 @@ void ovl_set_realfile(struct file *file, struct file *realfile)
>> ofe->realfile = realfile;
>> }
>>
>> -void *ovl_get_real_vmops(struct file *file)
>> +const struct vm_operations_struct *ovl_get_real_vmops(struct file *file)
>> {
>> struct ovl_file_entry *ofe = file->private_data;
>>
>> return ofe->vm_ops;
>> }
>>
>> -void ovl_set_real_vmops(struct file *file, void *vm_ops)
>> +void ovl_set_real_vmops(struct file *file,
>> + const struct vm_operations_struct *vm_ops)
>> {
>> struct ovl_file_entry *ofe = file->private_data;
>>
>> @@ -493,11 +502,104 @@ static int ovl_fsync(struct file *file, loff_t start, loff_t end, int datasync)
>> return ret;
>> }
>>
>> +vm_fault_t ovl_fault(struct vm_fault *vmf)
>> +{
>> + struct vm_area_struct *vma = vmf->vma;
>> + struct file *file = vma->vm_file;
>> + struct file *realfile;
>> + struct file *fpin, *tmp;
>> + struct inode *inode = file_inode(file);
>> + struct inode *realinode;
>> + const struct cred *old_cred;
>> + bool retry_allowed;
>> + vm_fault_t ret;
>> + int err = 0;
>> +
>> + if (fault_flag_check(vmf, FAULT_FLAG_TRIED)) {
>> + realfile = ovl_get_realfile(file);
>> +
>> + if (!ovl_has_upperdata(inode) ||
>> + realfile->f_inode != ovl_inode_upper(inode) ||
>> + !realfile->f_op->mmap)
>> + return VM_FAULT_SIGBUS;
>> +
>> + if (!ovl_get_real_vmops(file)) {
>> + old_cred = ovl_override_creds(inode->i_sb);
>> + err = call_mmap(realfile, vma);
>> + revert_creds(old_cred);
>> +
>> + vma->vm_file = file;
>> + if (err) {
>> + vma->vm_ops = &ovl_vm_ops;
>> + return VM_FAULT_SIGBUS;
>> + }
>> + ovl_set_real_vmops(file, vma->vm_ops);
>> + vma->vm_ops = &ovl_vm_ops;
>> + }
>> +
>> + retry_allowed = fault_flag_check(vmf, FAULT_FLAG_ALLOW_RETRY);
>> + if (retry_allowed)
>> + vma->vm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
>> + vma->vm_file = realfile;
>> + ret = ovl_get_real_vmops(file)->fault(vmf);
>> + vma->vm_file = file;
>> + if (retry_allowed)
>> + vma->vm_flags |= FAULT_FLAG_ALLOW_RETRY;
>> + return ret;
>> +
>> + } else {
>> + fpin = maybe_unlock_mmap_for_io(vmf, NULL);
>> + if (!fpin)
>> + return VM_FAULT_SIGBUS;
>> +
>> + ret = VM_FAULT_RETRY;
>> + if (!ovl_has_upperdata(inode)) {
>> + err = ovl_copy_up_with_data(file->f_path.dentry);
>> + if (err)
>> + goto out;
>> + }
>> +
>> + realinode = ovl_inode_realdata(inode);
>> + realfile = ovl_open_realfile(file, realinode);
>> + if (IS_ERR(realfile))
>> + goto out;
>> +
>> + tmp = ovl_get_realfile(file);
>> + ovl_set_realfile(file, realfile);
>> + fput(tmp);
>> +
>> +out:
>> + fput(fpin);
>> + return ret;
>> + }
>> +}
>
>
> Please add some documentation to explain the method used.
> Do we need to retry if real_vmops are already set?
>
Good catch, actually retry is not needed in that case.
Basically, we unlock(mmap_lock)->copy-up->open when
detecting no upper inode then retry fault operation.
However, we need to check fault retry flag carefully
for avoiding endless retry.
I'll add more explanation in v2.
---
cgxu
next prev parent reply other threads:[~2020-08-31 13:48 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-08-29 9:50 [RFC PATCH 0/3] ovl: stacked mmap for shared map Chengguang Xu
2020-08-29 9:50 ` [RFC PATCH 1/3] mm: mmap: export necessary functions for overlayfs' mmap Chengguang Xu
2020-08-29 9:51 ` [RFC PATCH 2/3] ovl: introduce struct ovl_file_entry Chengguang Xu
2020-08-29 9:51 ` [RFC PATCH 3/3] ovl: implement stacked mmap for shared map Chengguang Xu
2020-08-30 11:33 ` Amir Goldstein
2020-08-31 13:47 ` cgxu [this message]
2020-08-31 15:51 ` Amir Goldstein
2020-09-01 7:44 ` Miklos Szeredi
2020-09-01 13:20 ` cgxu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=e1e2c8f0-a3b8-0a3d-3093-6188b1a829f0@mykernel.net \
--to=cgxu519@mykernel.net \
--cc=akpm@linux-foundation.org \
--cc=amir73il@gmail.com \
--cc=linux-mm@kvack.org \
--cc=linux-unionfs@vger.kernel.org \
--cc=miklos@szeredi.hu \
--cc=riteshh@linux.ibm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).