ALSA: compress: Fix the mismatch size of struc between share lib(32bit) and kernel(64bit)
Subsystems (2)
PATCH 1/1 ALSA: compress: Fix the mismatch size of struc between share lib(32bit) and kernel(64bit)
The size of struct snd_compr_avail is 0x1c in 32bit kernel, while it is 0x20 in 64bit kernel 0x4 bytes added because of alignment. It is OK when 32bit kernel met 32bit user space. There exist stack corruption if 64bit kernel met 32bit user space, because the size of struct snd_compr_avail is 0x1c in 32bit user space which is smaller than it will get from kernel. The extra 4 bytes can corrupt the stack, and introduce unpredictable error.
--- include/uapi/sound/compress_offload.h | 1 + 1 files changed, 1 insertions(+), 0 deletions(-)
include/uapi/sound/compress_offload.h +1 -0
@@ -70,6 +70,7 @@ struct snd_compr_tstamp {__u32 pcm_frames;__u32 pcm_io_frames;__u32 sampling_rate; +__u32 reserved[1 ]; };/**
Signature · 1 line
1.7.1
At Mon, 09 Jun 2014 16:46:32 -0400,
Quoted 11 lines · Xiaoming
The size of struct
snd_compr_availis0x1cin 32bit kernel, while it is0x20in 64bit kernel0x4bytes added because of alignment. It is OK when 32bit kernel met 32bit user space. There exist stack corruption if 64bit kernel met 32bit user space, because the size of structsnd_compr_availis0x1cin 32bit user space which is smaller than it will get from kernel. The extra 4 bytes can corrupt the stack, and introduce unpredictable error.
Quoted 2 lines
Signed-off-by Zhang Dongxing <[email protected]>Signed-off-by xiaoming wang <[email protected]>
This would break the existing 32bit systems, so I don't think we can take this approach.
Either break the 64bit systems (which aren't deployed yet much, so far) by adding packed attribute, or implement 32/64 bit conversion in compat_ioctl fop.
thanks,
Takashi
Quoted 19 lines
--- include/uapi/sound/compress_offload.h | 1 + 1 files changed, 1 insertions(+), 0 deletions(-)include/uapi/sound/compress_offload.h +1
-0@@ -70,6 +70,7 @@ struct snd_compr_tstamp {__u32 pcm_frames;__u32 pcm_io_frames;__u32 sampling_rate; +__u32 reserved[1 ]; };/** Signature · 1 line
1.7.1
Quoted 16 lines · Takashi Iwai
At Mon, 09 Jun 2014 16:46:32 -0400,
Quoted 11 lines · Xiaoming
The size of struct
snd_compr_availis0x1cin 32bit kernel, while it is0x20in 64bit kernel0x4bytes added because of alignment. It is OK when 32bit kernel met 32bit user space. There exist stack corruption if 64bit kernel met 32bit user space, because the size of structsnd_compr_availis0x1cin 32bit user space which is smaller than it will get from kernel. The extra 4 bytes can corrupt the stack, and introduce unpredictable error.Quoted 2 lines
Signed-off-by Zhang Dongxing <[email protected]>Signed-off-by xiaoming wang <[email protected]>
Quoted 6 lines
This would break the existing 32bit systems, so I don't think we can take this approach.
Either break the 64bit systems (which aren't deployed yet much, so far) by adding packed attribute, or implement 32/64 bit conversion in
compat_ioctlfop.
I think former should be safe for now. Anyway we have only 1 driver using this in mainline so fallout shouldn't be widespread!
Signature · 1 line
~Vinod
Quoted 25 lines
thanks,
Takashi
Quoted 19 lines
--- include/uapi/sound/compress_offload.h | 1 + 1 files changed, 1 insertions(+), 0 deletions(-)include/uapi/sound/compress_offload.h +1
-0@@ -70,6 +70,7 @@ struct snd_compr_tstamp {__u32 pcm_frames;__u32 pcm_io_frames;__u32 sampling_rate; +__u32 reserved[1 ]; };/** Signature · 1 line
1.7.1
Dear Takashi
Quoted 27 lines
-----Original Message-----
From: Koul, Vinod
Sent: Monday, June 09, 2014 11:36 PM
To: Takashi Iwai
Cc: Wang, Xiaoming; Kp, Jeeja; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Zhang, Dongxing
Subject: Re: [PATCH] ALSA: compress: Fix the mismatch size of struc between share lib(32bit) and kernel(64bit)Quoted 15 lines · Takashi Iwai
At Mon, 09 Jun 2014 16:46:32 -0400,
Quoted 12 lines · Xiaoming
The size of struct
snd_compr_availis0x1cin 32bit kernel, while it is0x20in 64bit kernel0x4bytes added because of alignment. It is OK when 32bit kernel met 32bit user space.
There exist stack corruption if 64bit kernel met 32bit user space, because the size of structsnd_compr_availis0x1cin 32bit user space which is smaller than it will get from kernel. The extra 4 bytes can corrupt the stack, and introduce unpredictable error.Signed-off-by Zhang Dongxing <[email protected]>Signed-off-by xiaoming wang <[email protected]>
Quoted 8 lines
Quoted 6 lines
This would break the existing 32bit systems, so I don't think we can take this approach.
Either break the 64bit systems (which aren't deployed yet much, so far) by adding packed attribute, or implement 32/64 bit conversion in
compat_ioctlfop.I think former should be safe for now. Anyway we have only 1 driver using this in mainline so fallout shouldn't be widespread!
Can you accept we add __attribute__((packed, aligned(4))) for struct snd_compr_avail Which include the struct snd_compr_tstamp and fix the sizeof struct snd_compr_avail to 0x1c both in 32bit and 64 bit kernel?
Quoted 31 lines
--
~VinodQuoted 22 lines
thanks,
Takashi
Quoted 16 lines
--- include/uapi/sound/compress_offload.h | 1 + 1 files changed, 1 insertions(+), 0 deletions(-)include/uapi/sound/compress_offload.h +1
-0@@ -70,6 +70,7 @@ struct snd_compr_tstamp {__u32 pcm_frames;__u32 pcm_io_frames;__u32 sampling_rate; +__u32 reserved[1 ]; };Quoted 4 lines
Quoted 4 lines
include/uapi/sound/compress_offload.h +0
-1/** -- 1.7.1
--
Quoted 17 lines · Vinod Koul
Quoted 16 lines · Takashi Iwai
At Mon, 09 Jun 2014 16:46:32 -0400,
Quoted 13 lines · Xiaoming
The size of struct
snd_compr_availis0x1cin 32bit kernel, while it is0x20in 64bit kernel0x4bytes added because of alignment. It is OK when 32bit kernel met 32bit user space. There exist stack corruption if 64bit kernel met 32bit user space, because the size of structsnd_compr_availis0x1cin 32bit user space which is smaller than it will get from kernel. The extra 4 bytes can corrupt the stack, and introduce unpredictable error.Signed-off-by Zhang Dongxing <[email protected]>Signed-off-by xiaoming wang <[email protected]>
Quoted 8 lines
Quoted 6 lines
This would break the existing 32bit systems, so I don't think we can take this approach.
Either break the 64bit systems (which aren't deployed yet much, so far) by adding packed attribute, or implement 32/64 bit conversion in
compat_ioctlfop.I think former should be safe for now. Anyway we have only 1 driver using this in mainline so fallout shouldn't be widespread!
I guess since the driver was only merged for 3.16, we don't really have to worry about the ABI breakage anyway, but you can still use that approach if you have reason to stay compatible with existing used space built against out-of-tree drivers.
Anyway, if you use the __packed attribute, best apply it only to the individual __u64 member(s), not the entire struct, otherwise you might change user space programs in a subtle way when the alignment changes from 4 to 1 byte.
Arnd
Quoted 24 lines · Arnd Bergmann
Quoted 23 lines · Vinod Koul
Quoted 22 lines · Takashi Iwai
At Mon, 09 Jun 2014 16:46:32 -0400,
Quoted 11 lines · Xiaoming
The size of struct
snd_compr_availis0x1cin 32bit kernel, while it is0x20in 64bit kernel0x4bytes added because of alignment. It is OK when 32bit kernel met 32bit user space. There exist stack corruption if 64bit kernel met 32bit user space, because the size of structsnd_compr_availis0x1cin 32bit user space which is smaller than it will get from kernel. The extra 4 bytes can corrupt the stack, and introduce unpredictable error.Quoted 2 lines
Signed-off-by Zhang Dongxing <[email protected]>Signed-off-by xiaoming wang <[email protected]>This would break the existing 32bit systems, so I don't think we can take this approach.
Either break the 64bit systems (which aren't deployed yet much, so far) by adding packed attribute, or implement 32/64 bit conversion in
compat_ioctlfop.
Quoted 7 lines
Quoted 2 lines
I think former should be safe for now. Anyway we have only 1 driver using this in mainline so fallout shouldn't be widespread!
I guess since the driver was only merged for 3.16, we don't really have to worry about the ABI breakage anyway, but you can still use that approach if you have reason to stay compatible with existing used space built against out-of-tree drivers.
The qcom has on out of tree driver, but not sure if they have used on 64bit so they might not be impacted.
Quoted 4 lines
Anyway, if you use the
__packedattribute, best apply it only to the individual__u64member(s), not the entire struct, otherwise you might change user space programs in a subtle way when the alignment changes from 4 to 1 byte.
then wouldn't it make sense to call out the aligned as well to ensure that it is aligning to what we want. Then we should add aligned (4) everywhere as mostly we need 4 byte aligned here
Signature · 1 line
~Vinod
Quoted 8 lines · Vinod Koul
Quoted 4 lines
Anyway, if you use the
__packedattribute, best apply it only to the individual__u64member(s), not the entire struct, otherwise you might change user space programs in a subtle way when the alignment changes from 4 to 1 byte.then wouldn't it make sense to call out the aligned as well to ensure that it is aligning to what we want. Then we should add aligned (4) everywhere as mostly we need 4 byte aligned here
If you want to be explicit, then just mark the entire structure as attribute((packed,aligned(4))), not each individual member.
Arnd
Quoted 6 lines · Arnd Bergmann
Quoted 5 lines · Vinod Koul
Quoted 4 lines
Anyway, if you use the
__packedattribute, best apply it only to the individual__u64member(s), not the entire struct, otherwise you might change user space programs in a subtle way when the alignment changes from 4 to 1 byte.
Quoted 6 lines
Quoted 3 lines
then wouldn't it make sense to call out the aligned as well to ensure that it is aligning to what we want. Then we should add aligned (4) everywhere as mostly we need 4 byte aligned here
If you want to be explicit, then just mark the entire structure as attribute((packed,aligned(4))), not each individual member.
Oh yes, thats the idea. I will send out and also while reviewing found that packed was missing for few of other structs, so lets fix it now before 64bits systems proliferate and cause havoc :)
Signature · 1 line
~Vinod