| Message ID | 20241129055058.858940-2-willy@infradead.org (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | Remove accesses to page->index from ceph | expand |
On Fri, 2024-11-29 at 05:50 +0000, Matthew Wilcox (Oracle) wrote: > If the pages array contains encrypted pages, we cannot look at > page->index because that field is uninitialised. Instead, use the new > ceph_fscrypt_pagecache_folio() to get the pagecache folio and look at > the index of that. > > Fixes: 4de77f25fd85 (ceph: use osd_req_op_extent_osd_iter for netfs reads) This commit may be a better candidate for the fixes tag: d55207717ded ceph: add encryption support to writepage and writepages > Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> > --- > fs/ceph/addr.c | 5 ++++- > fs/ceph/crypto.h | 7 +++++++ > 2 files changed, 11 insertions(+), 1 deletion(-) > > diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c > index 4c82348fe1e6..284a6244fcdf 100644 > --- a/fs/ceph/addr.c > +++ b/fs/ceph/addr.c > @@ -1346,8 +1346,11 @@ static int ceph_writepages_start(struct address_space *mapping, > memset(data_pages + i, 0, > locked_pages * sizeof(*pages)); > } else { > + struct folio *folio; > + > BUG_ON(num_ops != req->r_num_ops); > - index = pages[i - 1]->index + 1; > + folio = ceph_fscrypt_pagecache_folio(pages[i - 1]); > + index = folio->index + 1; > /* request message now owns the pages array */ > pages = NULL; > } > diff --git a/fs/ceph/crypto.h b/fs/ceph/crypto.h > index 47e0c319fc68..7d75c4874aa8 100644 > --- a/fs/ceph/crypto.h > +++ b/fs/ceph/crypto.h > @@ -280,6 +280,13 @@ static inline struct page *ceph_fscrypt_pagecache_page(struct page *page) > } > #endif /* CONFIG_FS_ENCRYPTION */ > > +static inline struct folio *ceph_fscrypt_pagecache_folio(struct page *page) > +{ > + if (fscrypt_is_bounce_page(page)) > + page = fscrypt_pagecache_page(page); > + return page_folio(page); > +} > + > static inline loff_t ceph_fscrypt_page_offset(struct page *page) > { > return page_offset(ceph_fscrypt_pagecache_page(page)); Patch looks good though: Reviewed-by: Jeff Layton <jlayton@kernel.org>
On Fri, Nov 29, 2024 at 08:26:42AM -0500, Jeff Layton wrote: > On Fri, 2024-11-29 at 05:50 +0000, Matthew Wilcox (Oracle) wrote: > > If the pages array contains encrypted pages, we cannot look at > > page->index because that field is uninitialised. Instead, use the new > > ceph_fscrypt_pagecache_folio() to get the pagecache folio and look at > > the index of that. > > > > Fixes: 4de77f25fd85 (ceph: use osd_req_op_extent_osd_iter for netfs reads) > > This commit may be a better candidate for the fixes tag: > > d55207717ded ceph: add encryption support to writepage and writepages ... that's the commit i had intended to use. Not sure what I fat-fingered to make this happen.
diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c index 4c82348fe1e6..284a6244fcdf 100644 --- a/fs/ceph/addr.c +++ b/fs/ceph/addr.c @@ -1346,8 +1346,11 @@ static int ceph_writepages_start(struct address_space *mapping, memset(data_pages + i, 0, locked_pages * sizeof(*pages)); } else { + struct folio *folio; + BUG_ON(num_ops != req->r_num_ops); - index = pages[i - 1]->index + 1; + folio = ceph_fscrypt_pagecache_folio(pages[i - 1]); + index = folio->index + 1; /* request message now owns the pages array */ pages = NULL; } diff --git a/fs/ceph/crypto.h b/fs/ceph/crypto.h index 47e0c319fc68..7d75c4874aa8 100644 --- a/fs/ceph/crypto.h +++ b/fs/ceph/crypto.h @@ -280,6 +280,13 @@ static inline struct page *ceph_fscrypt_pagecache_page(struct page *page) } #endif /* CONFIG_FS_ENCRYPTION */ +static inline struct folio *ceph_fscrypt_pagecache_folio(struct page *page) +{ + if (fscrypt_is_bounce_page(page)) + page = fscrypt_pagecache_page(page); + return page_folio(page); +} + static inline loff_t ceph_fscrypt_page_offset(struct page *page) { return page_offset(ceph_fscrypt_pagecache_page(page));
If the pages array contains encrypted pages, we cannot look at page->index because that field is uninitialised. Instead, use the new ceph_fscrypt_pagecache_folio() to get the pagecache folio and look at the index of that. Fixes: 4de77f25fd85 (ceph: use osd_req_op_extent_osd_iter for netfs reads) Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> --- fs/ceph/addr.c | 5 ++++- fs/ceph/crypto.h | 7 +++++++ 2 files changed, 11 insertions(+), 1 deletion(-)