On 2023-10-10 18:39:40, Darrick J. Wong wrote:
> On Fri, Oct 06, 2023 at 08:49:19PM +0200, Andrey Albershteyn wrote:
> > - if (!args->value) {
> > + /*
> > + * We don't want to allocate memory for fs-verity Merkle tree blocks
> > + * (fs-verity descriptor is fine though). They will be stored in
> > + * underlying xfs_buf
> > + */
> > + if (!args->value && !xfs_verity_merkle_block(args)) {
>
> Hmm, why isn't this simply !(args->op_flags & XFS_DA_OP_BUFFER) ?
I check if args contains fs-verity block in multiple places so I
wrapped it into a function. Also, XFS_DA_OP_BUFFER only sets
args->bp. It doesn't affect where xfs_attr_copy_value copies the
value (this is changed with XBF_DOUBLE_ALLOC).
> > - if (args->attr_filter & XFS_ATTR_VERITY &&
> > - args->op_flags & XFS_DA_OP_BUFFER)
> > + if (xfs_verity_merkle_block(args))
> > flags = XBF_DOUBLE_ALLOC;
>
> Hmm, not sure what DOUBLE_ALLOC does, but I'll get there as I go
> backwards through the XFS patches...
>
> > +/*
> > + * fs-verity attribute name format
> > + *
> > + * Merkle tree blocks are stored under extended attributes of the inode. The
> > + * name of the attributes are offsets into merkle tree.
>
> Are these offsets byte offsets?
>
Byte offsets.
> > +++ b/fs/xfs/xfs_inode.h
> > @@ -342,7 +342,8 @@ static inline bool xfs_inode_has_large_extent_counts(struct xfs_inode *ip)
> > * inactivation completes, both flags will be cleared and the inode is a
> > * plain old IRECLAIMABLE inode.
> > */
> > -#define XFS_INACTIVATING (1 << 13)
> > +#define XFS_INACTIVATING (1 << 13)
> > +#define XFS_IVERITY_CONSTRUCTION (1 << 14) /* merkle tree construction */
>
> Under construction? Or already built?
Under construction.
> >
> > /* Quotacheck is running but inode has not been added to quota counts. */
> > #define XFS_IQUOTAUNCHECKED (1 << 14)
>
> These two iflags have the same definition.
Oops! Don't know how did I missed that, thanks!
> > diff --git a/fs/xfs/xfs_verity.c b/fs/xfs/xfs_verity.c
> > new file mode 100644
> > index 000000000000..a2db56974122
> > --- /dev/null
> > +++ b/fs/xfs/xfs_verity.c
> > @@ -0,0 +1,257 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * Copyright (C) 2023 Red Hat, Inc.
> > + */
> > +#include "xfs.h"
> > +#include "xfs_shared.h"
> > +#include "xfs_format.h"
> > +#include "xfs_da_format.h"
> > +#include "xfs_da_btree.h"
> > +#include "xfs_trans_resv.h"
> > +#include "xfs_mount.h"
> > +#include "xfs_inode.h"
> > +#include "xfs_attr.h"
> > +#include "xfs_verity.h"
> > +#include "xfs_bmap_util.h"
> > +#include "xfs_log_format.h"
> > +#include "xfs_trans.h"
> > +
> > +static int
>
> Hum, why isn't this ssize_t? That's what other IO functions return when
> returning the number of bytes acted upon.
The fs/verity prototype is with int, so I left it as int.
> > +static int
> > +xfs_drop_merkle_tree(
> > + struct xfs_inode *ip,
> > + u64 merkle_tree_size,
> > + u8 log_blocksize)
> > +{
> > + struct xfs_fsverity_merkle_key name;
> > + int error = 0, index;
> > + u64 offset = 0;
> > + struct xfs_da_args args = {
> > + .dp = ip,
> > + .whichfork = XFS_ATTR_FORK,
> > + .attr_filter = XFS_ATTR_VERITY,
> > + .namelen = sizeof(struct xfs_fsverity_merkle_key),
> > + /* NULL value make xfs_attr_set remove the attr */
> > + .value = NULL,
> > + };
> > +
> > + for (index = 1; offset < merkle_tree_size; index++) {
> > + xfs_fsverity_merkle_key_to_disk(&name, offset);
> > + args.name = (const uint8_t *)&name.merkleoff;
> > + args.attr_filter = XFS_ATTR_VERITY;
>
> Why do these two args. fields need to be reset every time through the
> loop?
Oh, yeah this should be ok. I will check it.
> > + error = xfs_attr_set(&args);
>
> Why is it ok to drop the error here?
>
My bad, will handle it.
> > + offset = index << log_blocksize;
>
> Hm, ok, the merkle key offsets /are/ byte offsets.
>
> Isn't this whole loop just:
>
> args.name = ...;
> args.attr_filter = ...;
> for (offset = 0; offset < merkle_tree_size; offset++) {
> xfs_fsverity_merkle_key_to_disk(&name, pos << log_blocksize);
> error = xfs_attr_set(&args);
> if (error)
> return error;
> }
>
> > +static int
> > +xfs_end_enable_verity(
> > + struct file *filp,
> > + const void *desc,
> > + size_t desc_size,
> > + u64 merkle_tree_size,
> > + u8 log_blocksize)
> > +{
> > + struct inode *inode = file_inode(filp);
> > + struct xfs_inode *ip = XFS_I(inode);
> > + struct xfs_mount *mp = ip->i_mount;
> > + struct xfs_trans *tp;
> > + struct xfs_da_args args = {
> > + .dp = ip,
> > + .whichfork = XFS_ATTR_FORK,
> > + .attr_filter = XFS_ATTR_VERITY,
> > + .attr_flags = XATTR_CREATE,
> > + .name = (const uint8_t *)XFS_VERITY_DESCRIPTOR_NAME,
> > + .namelen = XFS_VERITY_DESCRIPTOR_NAME_LEN,
> > + .value = (void *)desc,
> > + .valuelen = desc_size,
> > + };
> > + int error = 0;
> > +
> > + ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL));
> > +
> > + /* fs-verity failed, just cleanup */
> > + if (desc == NULL)
> > + goto out;
> > +
> > + error = xfs_attr_set(&args);
> > + if (error)
> > + goto out;
> > +
> > + /* Set fsverity inode flag */
> > + error = xfs_trans_alloc_inode(ip, &M_RES(mp)->tr_ichange,
> > + 0, 0, false, &tp);
> > + if (error)
> > + goto out;
> > +
> > + /*
> > + * Ensure that we've persisted the verity information before we enable
> > + * it on the inode and tell the caller we have sealed the inode.
> > + */
> > + ip->i_diflags2 |= XFS_DIFLAG2_VERITY;
> > +
> > + xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
> > + xfs_trans_set_sync(tp);
> > +
> > + error = xfs_trans_commit(tp);
> > + xfs_iunlock(ip, XFS_ILOCK_EXCL);
> > +
> > + if (!error)
> > + inode->i_flags |= S_VERITY;
> > +
> > +out:
> > + if (error)
> > + WARN_ON_ONCE(xfs_drop_merkle_tree(ip, merkle_tree_size,
> > + log_blocksize));
>
> (Don't WARNings panic some kernels?)
Not sure, but in case we fail to drop a tree (error already
happened) we probably want to know that something going on and tree
can not be removed. But I'm fine with removing this, not sure what
are the guidelines on WARNinigs.
> > +int
> > +xfs_read_merkle_tree_block(
> > + struct inode *inode,
> > + unsigned int pos,
> > + struct fsverity_block *block,
> > + unsigned long num_ra_pages)
> > +{
> > + struct xfs_inode *ip = XFS_I(inode);
> > + struct xfs_fsverity_merkle_key name;
> > + int error = 0;
> > + struct xfs_da_args args = {
> > + .dp = ip,
> > + .attr_filter = XFS_ATTR_VERITY,
> > + .namelen = sizeof(struct xfs_fsverity_merkle_key),
> > + };
> > + xfs_fsverity_merkle_key_to_disk(&name, pos);
> > + args.name = (const uint8_t *)&name.merkleoff;
> > +
> > + error = xfs_attr_get(&args);
> > + if (error)
> > + goto out;
> > +
> > + WARN_ON_ONCE(!args.valuelen);
> > +
> > + /* now we also want to get underlying xfs_buf */
> > + args.op_flags = XFS_DA_OP_BUFFER;
>
> If XFS_DA_OP_BUFFER returns the (bhold'd) xfs_buf containing the value,
> then ... can't we call xfs_attr_get once?
hmm, yeah, one call would be probably enough. The initial two calls
were here because fs/verity expected an error if attribute doesn't
exist. But with change to fs/verity/verify.c in patch 10 it's
probably fine to do everything just in one call. Thanks!
> Can the xfs_buf contents change after we drop the I{,O}LOCK?
By fs-verity? No
> Can users (or the kernel) change or add xattrs on a verity file?
>
> Are they allowed to move the file, and hence update the parent pointers?
Moving is probably fine but I haven't tested if it works fine with
parent pointers. But as they are xattrs I think it shouldn't be a
problem. I will check it.
> Or is the point of XBF_DOUBLE_ALLOC that we'll snapshot the attr data
> into the second half of the buffer, and that's what gets passed to
> fsverity core code?
Yes, this
> > +static int
> > +xfs_write_merkle_tree_block(
> > + struct inode *inode,
> > + const void *buf,
> > + u64 pos,
> > + unsigned int size)
> > +{
> > + struct xfs_inode *ip = XFS_I(inode);
> > + struct xfs_fsverity_merkle_key name;
> > + struct xfs_da_args args = {
> > + .dp = ip,
> > + .whichfork = XFS_ATTR_FORK,
> > + .attr_filter = XFS_ATTR_VERITY,
> > + .attr_flags = XATTR_CREATE,
>
> What happens if merkle tree fails midway through writing the blobs to
> the xattr tree? If they're not removed, then won't XATTR_CREATE here
> cause a second attempt to fail because there's a half-built tree?
If write fails during tree building fs-verity calls
xfs_end_enable_verity() which then goes into xfs_drop_merkle_tree()
(if(desc == NULL) case).
> > diff --git a/fs/xfs/xfs_verity.h b/fs/xfs/xfs_verity.h
> > new file mode 100644
> > index 000000000000..0f32fd212091
> > --- /dev/null
> > +++ b/fs/xfs/xfs_verity.h
> > @@ -0,0 +1,37 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * Copyright (C) 2022 Red Hat, Inc.
> > + */
> > +#ifndef __XFS_VERITY_H__
> > +#define __XFS_VERITY_H__
> > +
> > +#include "xfs.h"
> > +#include "xfs_da_format.h"
> > +#include "xfs_da_btree.h"
> > +#include <linux/fsverity.h>
> > +
> > +#define XFS_VERITY_DESCRIPTOR_NAME "verity_descriptor"
> > +#define XFS_VERITY_DESCRIPTOR_NAME_LEN 17
>
> Want to shorten this by one for u64 alignment? ;)
sure :)
> > +static inline bool
> > +xfs_verity_merkle_block(
> > + struct xfs_da_args *args)
> > +{
> > + if (!(args->attr_filter & XFS_ATTR_VERITY))
> > + return false;
> > +
> > + if (!(args->op_flags & XFS_DA_OP_BUFFER))
> > + return false;
> > +
> > + if (args->valuelen < 1024 || args->valuelen > PAGE_SIZE ||
> > + !is_power_of_2(args->valuelen))
> > + return false;
>
> Why do we check the valuelen here?
I use this function in multiple places to distinguish between
xfs_da_args which contains:
- fs-verity block
- fs-verity descriptor/any other attribute
Check for size is not necessary.
But this is one of the requirements on fs-verity blocks so I thought
it could be good candidate for one additional check. I'm fine with
removing this
> Also, if you're passing the xfs_buf out, I thought the buffer cache
> could handle weird sizes?
Not sure what you mean here
--
- Andrey
