Fix qcow2+luks corruption introduced by commit 8ac0f15f335
v4 cover Fix qcow2+luks corruption introduced by commit 8ac0f15f335
Commit 8ac0f15f335 accidently broke the COW of non changed areas of newly allocated clusters, when the write spans multiple clusters, and needs COW both prior and after the write.
This results in 'after' COW area being encrypted with wrong sector address, which render it corrupted.
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1745922
V2: grammar, spelling and code style fixes.
V3: more fixes after the review.
V4: addressed review comments from Max Reitz,
and futher refactored the qcow2_co_encrypt to just take full host and guest offset which simplifies everything.
Best regards,
Maxim Levitsky
Maxim Levitsky (3):
block/qcow2: refactoring of threaded encryption code block/qcow2: fix the corruption when rebasing luks encrypted files
qemu-iotests: Add test for bz #1745922
block/qcow2-cluster.c | 29 +++++++----- block/qcow2-threads.c | 62 ++++++++++++++++++++------ block/qcow2.c | 5 ++- block/qcow2.h | 8 ++-- tests/qemu-iotests/263 | 91 ++++++++++++++++++++++++++++++++++++++ tests/qemu-iotests/263.out | 40 +++++++++++++++++ tests/qemu-iotests/group | 1 + 7 files changed, 205 insertions(+), 31 deletions(-) create mode 100755 tests/qemu-iotests/263 create mode 100644 tests/qemu-iotests/263.out
Signature · 1 line
2.17.2
PATCH v4 1/3 block/qcow2: refactoring of threaded encryption code
This commit tries to clarify few function arguments, and add comments describing the encrypt/decrypt interface
--- block/qcow2-cluster.c | 9 ++++--- block/qcow2-threads.c | 62 ++++++++++++++++++++++++++++++++++--------- block/qcow2.c | 5 ++-- block/qcow2.h | 8 +++--- 4 files changed, 61 insertions(+), 23 deletions(-)
block/qcow2-cluster.c +5 -4
@@ -463,8 +463,8 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs, } static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, -uint64_t src_cluster_offset, -uint64_t cluster_offset, +uint64_t guest_cluster_offset, +uint64_t host_cluster_offset,unsigned offset_in_cluster,uint8_t *buffer,unsigned bytes)
@@ -474,8 +474,9 @@ static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs,assert ((offset_in_cluster & ~BDRV_SECTOR_MASK ) ==0 );assert ((bytes & ~BDRV_SECTOR_MASK ) ==0 );assert (s->crypto); -if (qcow2_co_encrypt (bs, cluster_offset, - src_cluster_offset + offset_in_cluster, +if (qcow2_co_encrypt (bs, + host_cluster_offset + offset_in_cluster, + guest_cluster_offset + offset_in_cluster, buffer, bytes) <0 ) {return false; }
block/qcow2-threads.c +49 -13
82 lines not on this page · raw message
block/qcow2.c +3 -2
@@ -2065,7 +2065,8 @@ static coroutine_fn int qcow2_co_preadv_part(BlockDriverState *bs,assert ((offset & (BDRV_SECTOR_SIZE -1 )) ==0 );assert ((cur_bytes & (BDRV_SECTOR_SIZE -1 )) ==0 ); -if (qcow2_co_decrypt (bs, cluster_offset, offset, +if (qcow2_co_decrypt (bs, cluster_offset + offset_in_cluster, + offset, cluster_data, cur_bytes) <0 ) { ret = -EIO ; goto fail;
@@ -2284,7 +2285,7 @@ static coroutine_fn int qcow2_co_pwritev_part(qemu_iovec_to_buf (qiov, qiov_offset + bytes_done, cluster_data, cur_bytes); -if (qcow2_co_encrypt (bs, cluster_offset, offset, +if (qcow2_co_encrypt (bs, cluster_offset + offset_in_cluster, offset, cluster_data, cur_bytes) <0 ) { ret = -EIO ; goto out_unlocked;
block/qcow2.h +4 -4
@@ -758,10 +758,10 @@ ssize_t coroutine_fn qcow2_co_decompress(BlockDriverState *bs, void *dest, size_t dest_size, const void *src, size_t src_size); int coroutine_fn -qcow2_co_encrypt (BlockDriverState *bs,uint64_t file_cluster_offset, -uint64_t offset,void *buf,size_t len); +qcow2_co_encrypt (BlockDriverState *bs,uint64_t host_offset, +uint64_t guest_offset,void *buf,size_t len); int coroutine_fn -qcow2_co_decrypt (BlockDriverState *bs,uint64_t file_cluster_offset, -uint64_t offset,void *buf,size_t len); +qcow2_co_decrypt (BlockDriverState *bs,uint64_t host_offset, +uint64_t guest_offset,void *buf,size_t len);#endif
Signature · 1 line
2.17.2
Re: [Qemu-devel] [PATCH v4 1/3] block/qcow2: refactoring of threaded encryption code
Quoted 27 lines · Maxim Levitsky
This commit tries to clarify few function arguments, and add comments describing the encrypt/decrypt interface
Signed-off-by Maxim Levitsky <[email protected]>--- block/qcow2-cluster.c | 9 ++++--- block/qcow2-threads.c | 62 ++++++++++++++++++++++++++++++++++--------- block/qcow2.c | 5 ++-- block/qcow2.h | 8 +++--- 4 files changed, 61 insertions(+), 23 deletions(-)block/qcow2-cluster.c +2
-2@@ -463,8 +463,8 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs, } static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, -uint64_t src_cluster_offset, -uint64_t cluster_offset, +uint64_t guest_cluster_offset, +uint64_t host_cluster_offset,unsigned offset_in_cluster,uint8_t *buffer,unsigned bytes)
block/qcow2-cluster.c static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, · line 479
Quoted 8 lines
block/qcow2-cluster.c +3
-2@@ -474,8 +474,9 @@ static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs,assert ((offset_in_cluster & ~BDRV_SECTOR_MASK ) ==0 );assert ((bytes & ~BDRV_SECTOR_MASK ) ==0 );assert (s->crypto); - if (qcow2_co_encrypt(bs, cluster_offset, - src_cluster_offset + offset_in_cluster, + if (qcow2_co_encrypt(bs, + host_cluster_offset + offset_in_cluster, + guest_cluster_offset + offset_in_cluster,
oops, seems you accidentally fixed the bug, which you are going to fix in the next patch, as now correct offsets are given to qcow2_co_encrypt :)
and next patch no is a simple no-logic-change refactoring, so at least commit message should be changed.
Quoted 15 lines
block/qcow2-cluster.c +0
-0buffer, bytes) < 0) {return false; }block/qcow2-threads.c +2
-2@@ -234,15 +234,15 @@ static int qcow2_encdec_pool_func(void *opaque) } static int coroutine_fn -qcow2_co_encdec (BlockDriverState *bs,uint64_t file_cluster_offset, -uint64_t offset,void *buf,size_t len,Qcow2EncDecFunc func) +qcow2_co_encdec(BlockDriverState *bs, uint64_t host_offset, + uint64_t guest_offset, void *buf, size_t len,
block/qcow2-threads.c static int qcow2_encdec_pool_func(void *opaque) · line 245
Quoted 10 lines
block/qcow2-threads.c +3
-3+ Qcow2EncDecFunc func) {BDRVQcow2State *s = bs->opaque; +Qcow2EncDecData arg = { .block = s->crypto, - .offset = s->crypt_physical_offset ? - file_cluster_offset +offset_into_cluster (s, offset) : - offset, + .offset = s->crypt_physical_offset ? host_offset : guest_offset,
Nice change. Obviously original design was worse.
Quoted 109 lines
block/qcow2-threads.c +44
-8.buf = buf, .len = len, .func = func,@@ -251,18 +251,54 @@ qcow2_co_encdec(BlockDriverState *bs, uint64_t file_cluster_offset,return qcow2_co_process (bs, qcow2_encdec_pool_func, &arg); } + +/* +* qcow2_co_encrypt() +* +* Encrypts one or more contiguous aligned sectors +* +* @host_offset - underlying storage offset of the first sector of the +* data to be encrypted + +* @guest_offset - guest (virtual) offset of the first sector of the +* data to be encrypted +* +* @buf - buffer with the data to encrypt, that after encryption +* will be written to the underlying storage device at +* @host_offset +* +* @len - length of the buffer (must be a BDRV_SECTOR_SIZE multiple) +* +* Depending on the encryption method, @host_cluster_offset and/or @guest_offset +* may be used for generating the initialization vector for +* encryption. +* +* Note that while the whole range must be aligned on sectors, it +* does not have to be aligned on clusters and can also cross cluster +* boundaries +* +*/ int coroutine_fn -qcow2_co_encrypt (BlockDriverState *bs,uint64_t file_cluster_offset, -uint64_t offset,void *buf,size_t len) +qcow2_co_encrypt (BlockDriverState *bs,uint64_t host_offset, +uint64_t guest_offset,void *buf,size_t len) { -return qcow2_co_encdec (bs, file_cluster_offset, offset, buf, len, - qcrypto_block_encrypt); +return qcow2_co_encdec (bs, host_offset, guest_offset, buf, len, + qcrypto_block_encrypt); } + +/* +* qcow2_co_decrypt() +* +* Decrypts one or more contiguous aligned sectors +* Similar to qcow2_co_encrypt +* +*/ + int coroutine_fn -qcow2_co_decrypt (BlockDriverState *bs,uint64_t file_cluster_offset, -uint64_t offset,void *buf,size_t len) +qcow2_co_decrypt (BlockDriverState *bs,uint64_t host_offset, +uint64_t guest_offset,void *buf,size_t len) { -return qcow2_co_encdec (bs, file_cluster_offset, offset, buf, len, - qcrypto_block_decrypt); +return qcow2_co_encdec (bs, host_offset, guest_offset, buf, len, + qcrypto_block_decrypt); }block/qcow2.c +3
-2@@ -2065,7 +2065,8 @@ static coroutine_fn int qcow2_co_preadv_part(BlockDriverState *bs,assert ((offset & (BDRV_SECTOR_SIZE -1 )) ==0 );assert ((cur_bytes & (BDRV_SECTOR_SIZE -1 )) ==0 ); -if (qcow2_co_decrypt (bs, cluster_offset, offset, +if (qcow2_co_decrypt (bs, cluster_offset + offset_in_cluster, + offset, cluster_data, cur_bytes) <0 ) { ret = -EIO ; goto fail;@@ -2284,7 +2285,7 @@ static coroutine_fn int qcow2_co_pwritev_part(qemu_iovec_to_buf (qiov, qiov_offset + bytes_done, cluster_data, cur_bytes); -if (qcow2_co_encrypt (bs, cluster_offset, offset, +if (qcow2_co_encrypt (bs, cluster_offset + offset_in_cluster, offset, cluster_data, cur_bytes) <0 ) { ret = -EIO ; goto out_unlocked;block/qcow2.h +4
-4@@ -758,10 +758,10 @@ ssize_t coroutine_fn qcow2_co_decompress(BlockDriverState *bs, void *dest, size_t dest_size, const void *src, size_t src_size); int coroutine_fn -qcow2_co_encrypt (BlockDriverState *bs,uint64_t file_cluster_offset, -uint64_t offset,void *buf,size_t len); +qcow2_co_encrypt (BlockDriverState *bs,uint64_t host_offset, +uint64_t guest_offset,void *buf,size_t len); int coroutine_fn -qcow2_co_decrypt (BlockDriverState *bs,uint64_t file_cluster_offset, -uint64_t offset,void *buf,size_t len); +qcow2_co_decrypt (BlockDriverState *bs,uint64_t host_offset, +uint64_t guest_offset,void *buf,size_t len);#endif
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v4 1/3] block/qcow2: refactoring of threaded encryption code
Quoted 40 lines · Vladimir Sementsov-Ogievskiy
Quoted 27 lines · Maxim Levitsky
This commit tries to clarify few function arguments, and add comments describing the encrypt/decrypt interface
Signed-off-by Maxim Levitsky <[email protected]>--- block/qcow2-cluster.c | 9 ++++--- block/qcow2-threads.c | 62 ++++++++++++++++++++++++++++++++++--------- block/qcow2.c | 5 ++-- block/qcow2.h | 8 +++--- 4 files changed, 61 insertions(+), 23 deletions(-)block/qcow2-cluster.c +2
-2@@ -463,8 +463,8 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs, } static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, -uint64_t src_cluster_offset, -uint64_t cluster_offset, +uint64_t guest_cluster_offset, +uint64_t host_cluster_offset,unsigned offset_in_cluster,uint8_t *buffer,unsigned bytes)block/qcow2-cluster.c static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, · line 479
Quoted 8 lines
block/qcow2-cluster.c +3
-2@@ -474,8 +474,9 @@ static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs,assert ((offset_in_cluster & ~BDRV_SECTOR_MASK ) ==0 );assert ((bytes & ~BDRV_SECTOR_MASK ) ==0 );assert (s->crypto); - if (qcow2_co_encrypt(bs, cluster_offset, - src_cluster_offset + offset_in_cluster, + if (qcow2_co_encrypt(bs, + host_cluster_offset + offset_in_cluster, + guest_cluster_offset + offset_in_cluster,oops, seems you accidentally fixed the bug, which you are going to fix in the next patch, as now correct offsets are given to
qcow2_co_encrypt:)
Quoted 2 lines
and next patch no is a simple no-logic-change refactoring, so at least commit message should be changed.
Yep :-( I am trying my best in addition to fixing the bug, also clarify the area to avoid this from happening again.
What do you think that I fold these two patches together after all? Best regards,
Maxim Levitsky
Re: [Qemu-devel] [PATCH v4 1/3] block/qcow2: refactoring of threaded encryption code
Quoted 41 lines · Maxim Levitsky
Quoted 40 lines · Vladimir Sementsov-Ogievskiy
Quoted 27 lines · Maxim Levitsky
This commit tries to clarify few function arguments, and add comments describing the encrypt/decrypt interface
Signed-off-by Maxim Levitsky <[email protected]>--- block/qcow2-cluster.c | 9 ++++--- block/qcow2-threads.c | 62 ++++++++++++++++++++++++++++++++++--------- block/qcow2.c | 5 ++-- block/qcow2.h | 8 +++--- 4 files changed, 61 insertions(+), 23 deletions(-)block/qcow2-cluster.c +2
-2@@ -463,8 +463,8 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs, } static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, -uint64_t src_cluster_offset, -uint64_t cluster_offset, +uint64_t guest_cluster_offset, +uint64_t host_cluster_offset,unsigned offset_in_cluster,uint8_t *buffer,unsigned bytes)block/qcow2-cluster.c static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, · line 479
Quoted 8 lines
block/qcow2-cluster.c +3
-2@@ -474,8 +474,9 @@ static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs,assert ((offset_in_cluster & ~BDRV_SECTOR_MASK ) ==0 );assert ((bytes & ~BDRV_SECTOR_MASK ) ==0 );assert (s->crypto); - if (qcow2_co_encrypt(bs, cluster_offset, - src_cluster_offset + offset_in_cluster, + if (qcow2_co_encrypt(bs, + host_cluster_offset + offset_in_cluster, + guest_cluster_offset + offset_in_cluster,oops, seems you accidentally fixed the bug, which you are going to fix in the next patch, as now correct offsets are given to
qcow2_co_encrypt:)
Quoted 5 lines
Quoted 2 lines
and next patch no is a simple no-logic-change refactoring, so at least commit message should be changed.
Yep :-( I am trying my best in addition to fixing the bug, also clarify the area to avoid this from happening again.
And this is really great! I'm sorry that I've broken these things, we actually don't use encryption (don't ask me why I've implemented threaded encryption :), so, I hope you are not only damaged by my patches but also have some benefit.
Quoted 2 lines
What do you think that I fold these two patches together after all?
OK for me (but I'm not a maintainer).
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v4 1/3] block/qcow2: refactoring of threaded encryption code
In reply to Maxim LevitskyAm 13.09.2019 um 16:07 hat Maxim Levitsky geschrieben:
Quoted 41 lines · Vladimir Sementsov-Ogievskiy
Quoted 40 lines
Quoted 27 lines · Maxim Levitsky
This commit tries to clarify few function arguments, and add comments describing the encrypt/decrypt interface
Signed-off-by Maxim Levitsky <[email protected]>--- block/qcow2-cluster.c | 9 ++++--- block/qcow2-threads.c | 62 ++++++++++++++++++++++++++++++++++--------- block/qcow2.c | 5 ++-- block/qcow2.h | 8 +++--- 4 files changed, 61 insertions(+), 23 deletions(-)block/qcow2-cluster.c +2
-2@@ -463,8 +463,8 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs, } static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, -uint64_t src_cluster_offset, -uint64_t cluster_offset, +uint64_t guest_cluster_offset, +uint64_t host_cluster_offset,unsigned offset_in_cluster,uint8_t *buffer,unsigned bytes)block/qcow2-cluster.c static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, · line 479
Quoted 8 lines
block/qcow2-cluster.c +3
-2@@ -474,8 +474,9 @@ static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs,assert ((offset_in_cluster & ~BDRV_SECTOR_MASK ) ==0 );assert ((bytes & ~BDRV_SECTOR_MASK ) ==0 );assert (s->crypto); - if (qcow2_co_encrypt(bs, cluster_offset, - src_cluster_offset + offset_in_cluster, + if (qcow2_co_encrypt(bs, + host_cluster_offset + offset_in_cluster, + guest_cluster_offset + offset_in_cluster,oops, seems you accidentally fixed the bug, which you are going to fix in the next patch, as now correct offsets are given to
qcow2_co_encrypt:)
Quoted 7 lines
Quoted 2 lines
and next patch no is a simple no-logic-change refactoring, so at least commit message should be changed.
Yep :-( I am trying my best in addition to fixing the bug, also clarify the area to avoid this from happening again.
What do you think that I fold these two patches together after all?
No, just make sure that your refactoring patch is really just refactoring without semantic change, i.e. make sure to preserve the bug in this patch.
Maybe you should actually have two refactoring patches (this one without the addition of offset_in_cluster, and patch 2) and an additional one-liner for the actual fix.
Kevin
Re: [Qemu-devel] [PATCH v4 1/3] block/qcow2: refactoring of threaded encryption code
Quoted 54 lines · Kevin Wolf
Am 13.09.2019 um 16:07 hat Maxim Levitsky geschrieben:
Quoted 41 lines · Vladimir Sementsov-Ogievskiy
Quoted 40 lines
Quoted 27 lines · Maxim Levitsky
This commit tries to clarify few function arguments, and add comments describing the encrypt/decrypt interface
Signed-off-by Maxim Levitsky <[email protected]>--- block/qcow2-cluster.c | 9 ++++--- block/qcow2-threads.c | 62 ++++++++++++++++++++++++++++++++++--------- block/qcow2.c | 5 ++-- block/qcow2.h | 8 +++--- 4 files changed, 61 insertions(+), 23 deletions(-)block/qcow2-cluster.c +2
-2@@ -463,8 +463,8 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs, } static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, -uint64_t src_cluster_offset, -uint64_t cluster_offset, +uint64_t guest_cluster_offset, +uint64_t host_cluster_offset,unsigned offset_in_cluster,uint8_t *buffer,unsigned bytes)block/qcow2-cluster.c static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, · line 479
Quoted 8 lines
block/qcow2-cluster.c +3
-2@@ -474,8 +474,9 @@ static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs,assert ((offset_in_cluster & ~BDRV_SECTOR_MASK ) ==0 );assert ((bytes & ~BDRV_SECTOR_MASK ) ==0 );assert (s->crypto); - if (qcow2_co_encrypt(bs, cluster_offset, - src_cluster_offset + offset_in_cluster, + if (qcow2_co_encrypt(bs, + host_cluster_offset + offset_in_cluster, + guest_cluster_offset + offset_in_cluster,oops, seems you accidentally fixed the bug, which you are going to fix in the next patch, as now correct offsets are given to
qcow2_co_encrypt:)Quoted 7 lines
Quoted 2 lines
and next patch no is a simple no-logic-change refactoring, so at least commit message should be changed.
Yep :-( I am trying my best in addition to fixing the bug, also clarify the area to avoid this from happening again.
What do you think that I fold these two patches together after all?
No, just make sure that your refactoring patch is really just refactoring without semantic change, i.e. make sure to preserve the bug in this patch.
Quoted 5 lines
Maybe you should actually have two refactoring patches (this one without the addition of
offset_in_cluster, and patch 2) and an additional one-liner for the actual fix.Kevin
Let me do it simplier I'll just split it to one liner patch that fixes it and second patch that does all the refactoring.
Best regards,
Maxim Levitsky
Re: [Qemu-devel] [PATCH v4 1/3] block/qcow2: refactoring of threaded encryption code
Quoted 55 lines · Maxim Levitsky
Quoted 54 lines · Kevin Wolf
Am 13.09.2019 um 16:07 hat Maxim Levitsky geschrieben:
Quoted 41 lines · Vladimir Sementsov-Ogievskiy
Quoted 40 lines
Quoted 27 lines · Maxim Levitsky
This commit tries to clarify few function arguments, and add comments describing the encrypt/decrypt interface
Signed-off-by Maxim Levitsky <[email protected]>--- block/qcow2-cluster.c | 9 ++++--- block/qcow2-threads.c | 62 ++++++++++++++++++++++++++++++++++--------- block/qcow2.c | 5 ++-- block/qcow2.h | 8 +++--- 4 files changed, 61 insertions(+), 23 deletions(-)block/qcow2-cluster.c +2
-2@@ -463,8 +463,8 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs, } static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, -uint64_t src_cluster_offset, -uint64_t cluster_offset, +uint64_t guest_cluster_offset, +uint64_t host_cluster_offset,unsigned offset_in_cluster,uint8_t *buffer,unsigned bytes)block/qcow2-cluster.c static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, · line 479
Quoted 8 lines
block/qcow2-cluster.c +3
-2@@ -474,8 +474,9 @@ static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs,assert ((offset_in_cluster & ~BDRV_SECTOR_MASK ) ==0 );assert ((bytes & ~BDRV_SECTOR_MASK ) ==0 );assert (s->crypto); - if (qcow2_co_encrypt(bs, cluster_offset, - src_cluster_offset + offset_in_cluster, + if (qcow2_co_encrypt(bs, + host_cluster_offset + offset_in_cluster, + guest_cluster_offset + offset_in_cluster,oops, seems you accidentally fixed the bug, which you are going to fix in the next patch, as now correct offsets are given to
qcow2_co_encrypt:)Quoted 7 lines
Quoted 2 lines
and next patch no is a simple no-logic-change refactoring, so at least commit message should be changed.
Yep :-( I am trying my best in addition to fixing the bug, also clarify the area to avoid this from happening again.
What do you think that I fold these two patches together after all?
No, just make sure that your refactoring patch is really just refactoring without semantic change, i.e. make sure to preserve the bug in this patch.
Quoted 9 lines
Quoted 5 lines
Maybe you should actually have two refactoring patches (this one without the addition of
offset_in_cluster, and patch 2) and an additional one-liner for the actual fix.Kevin
Let me do it simplier I'll just split it to one liner patch that fixes it and second patch that does all the refactoring.
[me typing actually the same suggestion in parallel, but you were the first]
I think it's the best: firstly fix the bug in a simple patch and then refactor to make code better.
I expect something like simply s/cluster_offset/start_of_cluster(cluster_offset + offset_in_cluster) in qcow2_co_encrypt call from do_perform_cow_encrypt,
yes?
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v4 1/3] block/qcow2: refactoring of threaded encryption code
Am 13.09.2019 um 16:44 hat Vladimir Sementsov-Ogievskiy geschrieben:
Quoted 68 lines
Quoted 55 lines · Maxim Levitsky
Quoted 54 lines · Kevin Wolf
Am 13.09.2019 um 16:07 hat Maxim Levitsky geschrieben:
Quoted 41 lines · Vladimir Sementsov-Ogievskiy
Quoted 40 lines
Quoted 27 lines · Maxim Levitsky
This commit tries to clarify few function arguments, and add comments describing the encrypt/decrypt interface
Signed-off-by Maxim Levitsky <[email protected]>--- block/qcow2-cluster.c | 9 ++++--- block/qcow2-threads.c | 62 ++++++++++++++++++++++++++++++++++--------- block/qcow2.c | 5 ++-- block/qcow2.h | 8 +++--- 4 files changed, 61 insertions(+), 23 deletions(-)block/qcow2-cluster.c +2
-2@@ -463,8 +463,8 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs, } static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, -uint64_t src_cluster_offset, -uint64_t cluster_offset, +uint64_t guest_cluster_offset, +uint64_t host_cluster_offset,unsigned offset_in_cluster,uint8_t *buffer,unsigned bytes)block/qcow2-cluster.c static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, · line 479
Quoted 8 lines
block/qcow2-cluster.c +3
-2@@ -474,8 +474,9 @@ static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs,assert ((offset_in_cluster & ~BDRV_SECTOR_MASK ) ==0 );assert ((bytes & ~BDRV_SECTOR_MASK ) ==0 );assert (s->crypto); - if (qcow2_co_encrypt(bs, cluster_offset, - src_cluster_offset + offset_in_cluster, + if (qcow2_co_encrypt(bs, + host_cluster_offset + offset_in_cluster, + guest_cluster_offset + offset_in_cluster,oops, seems you accidentally fixed the bug, which you are going to fix in the next patch, as now correct offsets are given to
qcow2_co_encrypt:)Quoted 7 lines
Quoted 2 lines
and next patch no is a simple no-logic-change refactoring, so at least commit message should be changed.
Yep :-( I am trying my best in addition to fixing the bug, also clarify the area to avoid this from happening again.
What do you think that I fold these two patches together after all?
No, just make sure that your refactoring patch is really just refactoring without semantic change, i.e. make sure to preserve the bug in this patch.
Quoted 9 lines
Quoted 5 lines
Maybe you should actually have two refactoring patches (this one without the addition of
offset_in_cluster, and patch 2) and an additional one-liner for the actual fix.Kevin
Let me do it simplier I'll just split it to one liner patch that fixes it and second patch that does all the refactoring.
[me typing actually the same suggestion in parallel, but you were the first]
Quoted 7 lines
I think it's the best: firstly fix the bug in a simple patch and then refactor to make code better.
I expect something like simply s/cluster_offset/start_of_cluster(cluster_offset +
offset_in_cluster) inqcow2_co_encryptcall fromdo_perform_cow_encrypt,
yes?
Yes, I think that's the right fix.
Kevin
Re: [Qemu-devel] [PATCH v4 1/3] block/qcow2: refactoring of threaded encryption code
In reply to Maxim LevitskyAm 13.09.2019 um 16:37 hat Maxim Levitsky geschrieben:
Quoted 55 lines · Kevin Wolf
Quoted 54 lines
Am 13.09.2019 um 16:07 hat Maxim Levitsky geschrieben:
Quoted 41 lines · Vladimir Sementsov-Ogievskiy
Quoted 40 lines
Quoted 27 lines · Maxim Levitsky
This commit tries to clarify few function arguments, and add comments describing the encrypt/decrypt interface
Signed-off-by Maxim Levitsky <[email protected]>--- block/qcow2-cluster.c | 9 ++++--- block/qcow2-threads.c | 62 ++++++++++++++++++++++++++++++++++--------- block/qcow2.c | 5 ++-- block/qcow2.h | 8 +++--- 4 files changed, 61 insertions(+), 23 deletions(-)block/qcow2-cluster.c +2
-2@@ -463,8 +463,8 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs, } static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, -uint64_t src_cluster_offset, -uint64_t cluster_offset, +uint64_t guest_cluster_offset, +uint64_t host_cluster_offset,unsigned offset_in_cluster,uint8_t *buffer,unsigned bytes)block/qcow2-cluster.c static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, · line 479
Quoted 8 lines
block/qcow2-cluster.c +3
-2@@ -474,8 +474,9 @@ static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs,assert ((offset_in_cluster & ~BDRV_SECTOR_MASK ) ==0 );assert ((bytes & ~BDRV_SECTOR_MASK ) ==0 );assert (s->crypto); - if (qcow2_co_encrypt(bs, cluster_offset, - src_cluster_offset + offset_in_cluster, + if (qcow2_co_encrypt(bs, + host_cluster_offset + offset_in_cluster, + guest_cluster_offset + offset_in_cluster,oops, seems you accidentally fixed the bug, which you are going to fix in the next patch, as now correct offsets are given to
qcow2_co_encrypt:)Quoted 7 lines
Quoted 2 lines
and next patch no is a simple no-logic-change refactoring, so at least commit message should be changed.
Yep :-( I am trying my best in addition to fixing the bug, also clarify the area to avoid this from happening again.
What do you think that I fold these two patches together after all?
No, just make sure that your refactoring patch is really just refactoring without semantic change, i.e. make sure to preserve the bug in this patch.
Quoted 8 lines
Quoted 5 lines
Maybe you should actually have two refactoring patches (this one without the addition of
offset_in_cluster, and patch 2) and an additional one-liner for the actual fix.Kevin
Let me do it simplier I'll just split it to one liner patch that fixes it and second patch that does all the refactoring.
Fine with me, as long as the fix is kept separate from the refactoring.
Kevin
PATCH v4 2/3 block/qcow2: fix the corruption when rebasing luks encrypted files
In reply to Maxim LevitskyThis fixes subtle corruption introduced by luks threaded encryption in commit 8ac0f15f335
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1745922
The corruption happens when we do a write that
* writes to two or more unallocated clusters at once
* doesn't fully cover the first sector
* doesn't fully cover the last sector
In this case, when allocating the new clusters we COW both areas prior to the write and after the write, and we encrypt them.
The above mentioned commit accidentally made it so we encrypt the second COW area using the physical cluster offset of the first area.
Fix this by:
* Remove the offset_in_cluster parameter of do_perform_cow_encrypt, since it is misleading. That offset can be larger than cluster size currently.
Instead just add the start and the end COW area offsets to both host and guest offsets that do_perform_cow_encrypt receives.
* in do_perform_cow_encrypt, remove the cluster offset from the host_offset, and thus pass correctly to the qcow2_co_encrypt, the host cluster offset and full guest offset
In the bugreport that was triggered by rebasing a luks image to new, zero filled base, which lot of such writes, and causes some files with zero areas to contain garbage there instead.
But as described above it can happen elsewhere as well
--- block/qcow2-cluster.c | 30 +++++++++++++++++------------- 1 file changed, 17 insertions(+), 13 deletions(-)
block/qcow2-cluster.c +17 -13
@@ -463,21 +463,21 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs, } static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs, -uint64_t guest_cluster_offset, -uint64_t host_cluster_offset, -unsigned offset_in_cluster, +uint64_t guest_offset, +uint64_t host_offset,uint8_t *buffer,unsigned bytes) {if (bytes && bs->encrypted) {BDRVQcow2State *s = bs->opaque; -assert ((offset_in_cluster & ~BDRV_SECTOR_MASK ) ==0 ); -assert ((bytes & ~BDRV_SECTOR_MASK ) ==0 ); + +assert (QEMU_IS_ALIGNED (guest_offset,BDRV_SECTOR_SIZE )); +assert (QEMU_IS_ALIGNED (host_offset,BDRV_SECTOR_SIZE )); +assert (QEMU_IS_ALIGNED (bytes,BDRV_SECTOR_SIZE ));assert (s->crypto); -if (qcow2_co_encrypt (bs, - host_cluster_offset + offset_in_cluster, - guest_cluster_offset + offset_in_cluster, - buffer, bytes) <0 ) { + +if (qcow2_co_encrypt (bs, host_offset, guest_offset, + buffer, bytes) <0 ) {return false; } }
@@ -891,11 +891,15 @@ static int perform_cow(BlockDriverState *bs, QCowL2Meta *m)/* Encrypt the data if necessary before writing it */ if (bs->encrypted) { -if (!do_perform_cow_encrypt (bs, m->offset, m->alloc_offset, - start->offset, start_buffer, +if (!do_perform_cow_encrypt (bs, + m->offset + start->offset, + m->alloc_offset + start->offset, + start_buffer, start->nb_bytes) || - !do_perform_cow_encrypt (bs, m->offset, m->alloc_offset, - end->offset, end_buffer, end->nb_bytes)) { + !do_perform_cow_encrypt (bs, + m->offset + end->offset, + m->alloc_offset + end->offset, + end_buffer, end->nb_bytes)) { ret = -EIO ; goto fail; }
Signature · 1 line
2.17.2
PATCH v4 3/3 qemu-iotests: Add test for bz #1745922
In reply to Maxim Levitsky--- tests/qemu-iotests/263 | 91 ++++++++++++++++++++++++++++++++++++++ tests/qemu-iotests/263.out | 40 +++++++++++++++++ tests/qemu-iotests/group | 1 + 3 files changed, 132 insertions(+) create mode 100755 tests/qemu-iotests/263 create mode 100644 tests/qemu-iotests/263.out
tests/qemu-iotests/263 new file +91 -0
@@ -0,0 +1,91 @@ +#!/usr/bin/env bash +# +# Test encrypted write that crosses cluster boundary of two unallocated clusters +# Based on 188 +# +# Copyright (C) 2019 Red Hat, Inc. +# +# This program is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 2 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see <http://www.gnu.org/licenses/>. +# + +# creator [email protected] + +seq=`basename $0` +echo "QA output created by $seq" + +status=1 # failure is the default! + +_cleanup() +{ + _cleanup_test_img +} +trap "_cleanup; exit \$status" 0 1 2 3 15 + +# get standard environment, filters and checks +. ./common.rc +. ./common.filter + +_supported_fmt qcow2 +_supported_proto generic +_supported_os Linux + + +size=1M + +SECRET="secret,id=sec0,data=astrochicken" +QEMU_IO_OPTIONS=$QEMU_IO_OPTIONS_NO_FMT + + +_run_test() +{ + echo "== reading the whole image ==" + $QEMU_IO --object $SECRET -c "read -P 0 0 $size" --image-opts "$1" | _filter_qemu_io | _filter_testdir + + echo + echo "== write two 512 byte sectors on a cluster boundary ==" + $QEMU_IO --object $SECRET -c "write -P 0xAA 0xFE00 0x400" --image-opts "$1" | _filter_qemu_io | _filter_testdir + + echo + echo "== verify that the rest of the image is not changed ==" + $QEMU_IO --object $SECRET -c "read -P 0x00 0x00000 0xFE00" --image-opts "$1" | _filter_qemu_io | _filter_testdir + $QEMU_IO --object $SECRET -c "read -P 0xAA 0x0FE00 0x400" --image-opts "$1" | _filter_qemu_io | _filter_testdir + $QEMU_IO --object $SECRET -c "read -P 0x00 0x10200 0xEFE00" --image-opts "$1" | _filter_qemu_io | _filter_testdir + +} + + +echo +echo "testing LUKS qcow2 encryption" +echo + +_make_test_img --object $SECRET -o "encrypt.format=luks,encrypt.key-secret=sec0,encrypt.iter-time=10,cluster_size=64K" $size +_run_test "driver=$IMGFMT,encrypt.key-secret=sec0,file.filename=$TEST_IMG" +_cleanup_test_img + +echo +echo "testing legacy AES qcow2 encryption" +echo + + +_make_test_img --object $SECRET -o "encrypt.format=aes,encrypt.key-secret=sec0,cluster_size=64K" $size +_run_test "driver=$IMGFMT,encrypt.key-secret=sec0,file.filename=$TEST_IMG" +_cleanup_test_img + + + +# success, all done +echo "*** done" +rm -f $seq.full +status=0
tests/qemu-iotests/263.out new file +40 -0
@@ -0,0 +1,40 @@ +QA output created by 263 + +testing LUKS qcow2 encryption + +Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=1048576 encrypt.format=luks encrypt.key-secret=sec0 encrypt.iter-time=10 +== reading the whole image == +read 1048576/1048576 bytes at offset 0 +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) + +== write two 512 byte sectors on a cluster boundary == +wrote 1024/1024 bytes at offset 65024 +1 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) + +== verify that the rest of the image is not changed == +read 65024/65024 bytes at offset 0 +63.500 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) +read 1024/1024 bytes at offset 65024 +1 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) +read 982528/982528 bytes at offset 66048 +959.500 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) + +testing legacy AES qcow2 encryption + +Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=1048576 encrypt.format=aes encrypt.key-secret=sec0 +== reading the whole image == +read 1048576/1048576 bytes at offset 0 +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) + +== write two 512 byte sectors on a cluster boundary == +wrote 1024/1024 bytes at offset 65024 +1 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) + +== verify that the rest of the image is not changed == +read 65024/65024 bytes at offset 0 +63.500 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) +read 1024/1024 bytes at offset 65024 +1 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) +read 982528/982528 bytes at offset 66048 +959.500 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) +*** done
tests/qemu-iotests/group +1 -0
@@ -274,3 +274,4 @@ 257 rw 258 rw quick 262 rw quick migration +263 rw quick
Signature · 1 line
2.17.2