| Message ID | 20191209200055.204040-1-ebiggers@kernel.org (mailing list archive) |
|---|---|
| State | Accepted |
| Headers | show |
| Series | f2fs: don't keep META_MAPPING pages used for moving verity file blocks | expand |
On 12/09, Eric Biggers wrote: > From: Eric Biggers <ebiggers@google.com> > > META_MAPPING is used to move blocks for both encrypted and verity files. > So the META_MAPPING invalidation condition in do_checkpoint() should > consider verity too, not just encrypt. > > Signed-off-by: Eric Biggers <ebiggers@google.com> > --- > fs/f2fs/checkpoint.c | 6 +++--- > 1 file changed, 3 insertions(+), 3 deletions(-) > > diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c > index ffdaba0c55d29..44e84ac5c9411 100644 > --- a/fs/f2fs/checkpoint.c > +++ b/fs/f2fs/checkpoint.c > @@ -1509,10 +1509,10 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) > f2fs_wait_on_all_pages_writeback(sbi); > > /* > - * invalidate intermediate page cache borrowed from meta inode > - * which are used for migration of encrypted inode's blocks. > + * invalidate intermediate page cache borrowed from meta inode which are > + * used for migration of encrypted or verity inode's blocks. > */ > - if (f2fs_sb_has_encrypt(sbi)) > + if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi)) Do we need f2fs_post_read_required() aligned to the condition of move_data_block()? > invalidate_mapping_pages(META_MAPPING(sbi), > MAIN_BLKADDR(sbi), MAX_BLKADDR(sbi) - 1); > > -- > 2.24.0.393.g34dc348eaf-goog
On Mon, Dec 09, 2019 at 02:28:28PM -0800, Jaegeuk Kim wrote: > On 12/09, Eric Biggers wrote: > > From: Eric Biggers <ebiggers@google.com> > > > > META_MAPPING is used to move blocks for both encrypted and verity files. > > So the META_MAPPING invalidation condition in do_checkpoint() should > > consider verity too, not just encrypt. > > > > Signed-off-by: Eric Biggers <ebiggers@google.com> > > --- > > fs/f2fs/checkpoint.c | 6 +++--- > > 1 file changed, 3 insertions(+), 3 deletions(-) > > > > diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c > > index ffdaba0c55d29..44e84ac5c9411 100644 > > --- a/fs/f2fs/checkpoint.c > > +++ b/fs/f2fs/checkpoint.c > > @@ -1509,10 +1509,10 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) > > f2fs_wait_on_all_pages_writeback(sbi); > > > > /* > > - * invalidate intermediate page cache borrowed from meta inode > > - * which are used for migration of encrypted inode's blocks. > > + * invalidate intermediate page cache borrowed from meta inode which are > > + * used for migration of encrypted or verity inode's blocks. > > */ > > - if (f2fs_sb_has_encrypt(sbi)) > > + if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi)) > > Do we need f2fs_post_read_required() aligned to the condition of > move_data_block()? > I think you're asking why verity files have to be moved via META_MAPPING? The reason is that we have to be super careful not to read pages of a verity file into its own address_space without doing the fs-verity data verification, as then unverified data would be available to userspace. In theory, F2FS's garbage collector could do the data verification. But it's tricky because ->i_verity_info may not have been set up yet. So it might be easiest to continue to treat verity files like encrypted files. - Eric
On 12/09, Eric Biggers wrote: > On Mon, Dec 09, 2019 at 02:28:28PM -0800, Jaegeuk Kim wrote: > > On 12/09, Eric Biggers wrote: > > > From: Eric Biggers <ebiggers@google.com> > > > > > > META_MAPPING is used to move blocks for both encrypted and verity files. > > > So the META_MAPPING invalidation condition in do_checkpoint() should > > > consider verity too, not just encrypt. > > > > > > Signed-off-by: Eric Biggers <ebiggers@google.com> > > > --- > > > fs/f2fs/checkpoint.c | 6 +++--- > > > 1 file changed, 3 insertions(+), 3 deletions(-) > > > > > > diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c > > > index ffdaba0c55d29..44e84ac5c9411 100644 > > > --- a/fs/f2fs/checkpoint.c > > > +++ b/fs/f2fs/checkpoint.c > > > @@ -1509,10 +1509,10 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) > > > f2fs_wait_on_all_pages_writeback(sbi); > > > > > > /* > > > - * invalidate intermediate page cache borrowed from meta inode > > > - * which are used for migration of encrypted inode's blocks. > > > + * invalidate intermediate page cache borrowed from meta inode which are > > > + * used for migration of encrypted or verity inode's blocks. > > > */ > > > - if (f2fs_sb_has_encrypt(sbi)) > > > + if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi)) > > > > Do we need f2fs_post_read_required() aligned to the condition of > > move_data_block()? > > > > I think you're asking why verity files have to be moved via META_MAPPING? The > reason is that we have to be super careful not to read pages of a verity file > into its own address_space without doing the fs-verity data verification, as > then unverified data would be available to userspace. > > In theory, F2FS's garbage collector could do the data verification. But it's > tricky because ->i_verity_info may not have been set up yet. So it might be > easiest to continue to treat verity files like encrypted files. I meant to ask why not just checking f2fs_post_read_required() here so that we could sync the check across multiple places. > > - Eric
On Mon, Dec 09, 2019 at 02:48:57PM -0800, Jaegeuk Kim wrote: > On 12/09, Eric Biggers wrote: > > On Mon, Dec 09, 2019 at 02:28:28PM -0800, Jaegeuk Kim wrote: > > > On 12/09, Eric Biggers wrote: > > > > From: Eric Biggers <ebiggers@google.com> > > > > > > > > META_MAPPING is used to move blocks for both encrypted and verity files. > > > > So the META_MAPPING invalidation condition in do_checkpoint() should > > > > consider verity too, not just encrypt. > > > > > > > > Signed-off-by: Eric Biggers <ebiggers@google.com> > > > > --- > > > > fs/f2fs/checkpoint.c | 6 +++--- > > > > 1 file changed, 3 insertions(+), 3 deletions(-) > > > > > > > > diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c > > > > index ffdaba0c55d29..44e84ac5c9411 100644 > > > > --- a/fs/f2fs/checkpoint.c > > > > +++ b/fs/f2fs/checkpoint.c > > > > @@ -1509,10 +1509,10 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) > > > > f2fs_wait_on_all_pages_writeback(sbi); > > > > > > > > /* > > > > - * invalidate intermediate page cache borrowed from meta inode > > > > - * which are used for migration of encrypted inode's blocks. > > > > + * invalidate intermediate page cache borrowed from meta inode which are > > > > + * used for migration of encrypted or verity inode's blocks. > > > > */ > > > > - if (f2fs_sb_has_encrypt(sbi)) > > > > + if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi)) > > > > > > Do we need f2fs_post_read_required() aligned to the condition of > > > move_data_block()? > > > > > > > I think you're asking why verity files have to be moved via META_MAPPING? The > > reason is that we have to be super careful not to read pages of a verity file > > into its own address_space without doing the fs-verity data verification, as > > then unverified data would be available to userspace. > > > > In theory, F2FS's garbage collector could do the data verification. But it's > > tricky because ->i_verity_info may not have been set up yet. So it might be > > easiest to continue to treat verity files like encrypted files. > > I meant to ask why not just checking f2fs_post_read_required() here so that we > could sync the check across multiple places. > We can't use f2fs_post_read_required() here because f2fs_post_read_required() deals with a specific inode, but do_checkpoint() is dealing with the entire filesystem. It's checking whether the filesystem might have *any* files that were moved via META_MAPPING. - Eric
On 12/09, Eric Biggers wrote: > On Mon, Dec 09, 2019 at 02:48:57PM -0800, Jaegeuk Kim wrote: > > On 12/09, Eric Biggers wrote: > > > On Mon, Dec 09, 2019 at 02:28:28PM -0800, Jaegeuk Kim wrote: > > > > On 12/09, Eric Biggers wrote: > > > > > From: Eric Biggers <ebiggers@google.com> > > > > > > > > > > META_MAPPING is used to move blocks for both encrypted and verity files. > > > > > So the META_MAPPING invalidation condition in do_checkpoint() should > > > > > consider verity too, not just encrypt. > > > > > > > > > > Signed-off-by: Eric Biggers <ebiggers@google.com> > > > > > --- > > > > > fs/f2fs/checkpoint.c | 6 +++--- > > > > > 1 file changed, 3 insertions(+), 3 deletions(-) > > > > > > > > > > diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c > > > > > index ffdaba0c55d29..44e84ac5c9411 100644 > > > > > --- a/fs/f2fs/checkpoint.c > > > > > +++ b/fs/f2fs/checkpoint.c > > > > > @@ -1509,10 +1509,10 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) > > > > > f2fs_wait_on_all_pages_writeback(sbi); > > > > > > > > > > /* > > > > > - * invalidate intermediate page cache borrowed from meta inode > > > > > - * which are used for migration of encrypted inode's blocks. > > > > > + * invalidate intermediate page cache borrowed from meta inode which are > > > > > + * used for migration of encrypted or verity inode's blocks. > > > > > */ > > > > > - if (f2fs_sb_has_encrypt(sbi)) > > > > > + if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi)) > > > > > > > > Do we need f2fs_post_read_required() aligned to the condition of > > > > move_data_block()? > > > > > > > > > > I think you're asking why verity files have to be moved via META_MAPPING? The > > > reason is that we have to be super careful not to read pages of a verity file > > > into its own address_space without doing the fs-verity data verification, as > > > then unverified data would be available to userspace. > > > > > > In theory, F2FS's garbage collector could do the data verification. But it's > > > tricky because ->i_verity_info may not have been set up yet. So it might be > > > easiest to continue to treat verity files like encrypted files. > > > > I meant to ask why not just checking f2fs_post_read_required() here so that we > > could sync the check across multiple places. > > > > We can't use f2fs_post_read_required() here because f2fs_post_read_required() > deals with a specific inode, but do_checkpoint() is dealing with the entire > filesystem. It's checking whether the filesystem might have *any* files that > were moved via META_MAPPING. Ah, I misread the code. Thank you for refreshing me. > > - Eric
On 2019/12/10 4:00, Eric Biggers wrote: > From: Eric Biggers <ebiggers@google.com> > > META_MAPPING is used to move blocks for both encrypted and verity files. > So the META_MAPPING invalidation condition in do_checkpoint() should > consider verity too, not just encrypt. > > Signed-off-by: Eric Biggers <ebiggers@google.com> Reviewed-by: Chao Yu <yuchao0@huawei.com> Thanks,
diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c index ffdaba0c55d29..44e84ac5c9411 100644 --- a/fs/f2fs/checkpoint.c +++ b/fs/f2fs/checkpoint.c @@ -1509,10 +1509,10 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) f2fs_wait_on_all_pages_writeback(sbi); /* - * invalidate intermediate page cache borrowed from meta inode - * which are used for migration of encrypted inode's blocks. + * invalidate intermediate page cache borrowed from meta inode which are + * used for migration of encrypted or verity inode's blocks. */ - if (f2fs_sb_has_encrypt(sbi)) + if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi)) invalidate_mapping_pages(META_MAPPING(sbi), MAIN_BLKADDR(sbi), MAX_BLKADDR(sbi) - 1);