2020-07-27 18:55:44

by Anthony Yznaga

[permalink] [raw]
Subject: [RFC PATCH 5/5] mm: introduce MADV_DOEXEC

madvise MADV_DOEXEC preserves a memory range across exec. Initially
only supported for non-executable, non-stack, anonymous memory.
MADV_DONTEXEC reverts the effect of a previous MADV_DOXEXEC call and
undoes the preservation of the range. After a successful exec call,
the behavior of all ranges reverts to MADV_DONTEXEC.

Signed-off-by: Steve Sistare <[email protected]>
Signed-off-by: Anthony Yznaga <[email protected]>
---
include/uapi/asm-generic/mman-common.h | 3 +++
mm/madvise.c | 25 +++++++++++++++++++++++++
2 files changed, 28 insertions(+)

diff --git a/include/uapi/asm-generic/mman-common.h b/include/uapi/asm-generic/mman-common.h
index f94f65d429be..7c5f616b28f7 100644
--- a/include/uapi/asm-generic/mman-common.h
+++ b/include/uapi/asm-generic/mman-common.h
@@ -72,6 +72,9 @@
#define MADV_COLD 20 /* deactivate these pages */
#define MADV_PAGEOUT 21 /* reclaim these pages */

+#define MADV_DOEXEC 22 /* do inherit across exec */
+#define MADV_DONTEXEC 23 /* don't inherit across exec */
+
/* compatibility flags */
#define MAP_FILE 0

diff --git a/mm/madvise.c b/mm/madvise.c
index dd1d43cf026d..b447fa748649 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -103,6 +103,26 @@ static long madvise_behavior(struct vm_area_struct *vma,
case MADV_KEEPONFORK:
new_flags &= ~VM_WIPEONFORK;
break;
+ case MADV_DOEXEC:
+ /*
+ * MADV_DOEXEC is only supported on private, non-executable,
+ * non-stack anonymous memory and if the VM_EXEC_KEEP flag
+ * is available.
+ */
+ if (!VM_EXEC_KEEP || vma->vm_file || vma->vm_flags & (VM_EXEC|VM_SHARED|VM_STACK)) {
+ error = -EINVAL;
+ goto out;
+ }
+ new_flags |= (new_flags & ~VM_MAYEXEC) | VM_EXEC_KEEP;
+ break;
+ case MADV_DONTEXEC:
+ if (!VM_EXEC_KEEP) {
+ error = -EINVAL;
+ goto out;
+ }
+ if (new_flags & VM_EXEC_KEEP)
+ new_flags |= (new_flags & ~VM_EXEC_KEEP) | VM_MAYEXEC;
+ break;
case MADV_DONTDUMP:
new_flags |= VM_DONTDUMP;
break;
@@ -983,6 +1003,8 @@ static int madvise_inject_error(int behavior,
case MADV_SOFT_OFFLINE:
case MADV_HWPOISON:
#endif
+ case MADV_DOEXEC:
+ case MADV_DONTEXEC:
return true;

default:
@@ -1037,6 +1059,9 @@ static int madvise_inject_error(int behavior,
* MADV_DONTDUMP - the application wants to prevent pages in the given range
* from being included in its core dump.
* MADV_DODUMP - cancel MADV_DONTDUMP: no longer exclude from core dump.
+ * MADV_DOEXEC - On exec, preserve and duplicate this area in the new process
+ * if the new process allows it.
+ * MADV_DONTEXEC - Undo the effect of MADV_DOEXEC.
*
* return values:
* zero - success
--
1.8.3.1


2020-07-28 13:24:13

by Kirill Tkhai

[permalink] [raw]
Subject: Re: [RFC PATCH 5/5] mm: introduce MADV_DOEXEC

On 27.07.2020 20:11, Anthony Yznaga wrote:
> madvise MADV_DOEXEC preserves a memory range across exec. Initially
> only supported for non-executable, non-stack, anonymous memory.
> MADV_DONTEXEC reverts the effect of a previous MADV_DOXEXEC call and
> undoes the preservation of the range. After a successful exec call,
> the behavior of all ranges reverts to MADV_DONTEXEC.
>
> Signed-off-by: Steve Sistare <[email protected]>
> Signed-off-by: Anthony Yznaga <[email protected]>
> ---
> include/uapi/asm-generic/mman-common.h | 3 +++
> mm/madvise.c | 25 +++++++++++++++++++++++++
> 2 files changed, 28 insertions(+)
>
> diff --git a/include/uapi/asm-generic/mman-common.h b/include/uapi/asm-generic/mman-common.h
> index f94f65d429be..7c5f616b28f7 100644
> --- a/include/uapi/asm-generic/mman-common.h
> +++ b/include/uapi/asm-generic/mman-common.h
> @@ -72,6 +72,9 @@
> #define MADV_COLD 20 /* deactivate these pages */
> #define MADV_PAGEOUT 21 /* reclaim these pages */
>
> +#define MADV_DOEXEC 22 /* do inherit across exec */
> +#define MADV_DONTEXEC 23 /* don't inherit across exec */
> +
> /* compatibility flags */
> #define MAP_FILE 0
>
> diff --git a/mm/madvise.c b/mm/madvise.c
> index dd1d43cf026d..b447fa748649 100644
> --- a/mm/madvise.c
> +++ b/mm/madvise.c
> @@ -103,6 +103,26 @@ static long madvise_behavior(struct vm_area_struct *vma,
> case MADV_KEEPONFORK:
> new_flags &= ~VM_WIPEONFORK;
> break;
> + case MADV_DOEXEC:

For me MADV_KEEPONEXEC sounds better as it's symmetric to MADV_KEEPONFORK.

> + /*
> + * MADV_DOEXEC is only supported on private, non-executable,
> + * non-stack anonymous memory and if the VM_EXEC_KEEP flag
> + * is available.
> + */
> + if (!VM_EXEC_KEEP || vma->vm_file || vma->vm_flags & (VM_EXEC|VM_SHARED|VM_STACK)) {
> + error = -EINVAL;
> + goto out;
> + }
> + new_flags |= (new_flags & ~VM_MAYEXEC) | VM_EXEC_KEEP;
> + break;
> + case MADV_DONTEXEC:
> + if (!VM_EXEC_KEEP) {
> + error = -EINVAL;
> + goto out;
> + }
> + if (new_flags & VM_EXEC_KEEP)
> + new_flags |= (new_flags & ~VM_EXEC_KEEP) | VM_MAYEXEC;
> + break;
> case MADV_DONTDUMP:
> new_flags |= VM_DONTDUMP;
> break;
> @@ -983,6 +1003,8 @@ static int madvise_inject_error(int behavior,
> case MADV_SOFT_OFFLINE:
> case MADV_HWPOISON:
> #endif
> + case MADV_DOEXEC:
> + case MADV_DONTEXEC:
> return true;
>
> default:
> @@ -1037,6 +1059,9 @@ static int madvise_inject_error(int behavior,
> * MADV_DONTDUMP - the application wants to prevent pages in the given range
> * from being included in its core dump.
> * MADV_DODUMP - cancel MADV_DONTDUMP: no longer exclude from core dump.
> + * MADV_DOEXEC - On exec, preserve and duplicate this area in the new process
> + * if the new process allows it.
> + * MADV_DONTEXEC - Undo the effect of MADV_DOEXEC.
> *
> * return values:
> * zero - success
>

2020-07-28 14:07:41

by Steven Sistare

[permalink] [raw]
Subject: Re: [RFC PATCH 5/5] mm: introduce MADV_DOEXEC

On 7/28/2020 9:22 AM, Kirill Tkhai wrote:
> On 27.07.2020 20:11, Anthony Yznaga wrote:
>> madvise MADV_DOEXEC preserves a memory range across exec. Initially
>> only supported for non-executable, non-stack, anonymous memory.
>> MADV_DONTEXEC reverts the effect of a previous MADV_DOXEXEC call and
>> undoes the preservation of the range. After a successful exec call,
>> the behavior of all ranges reverts to MADV_DONTEXEC.
>>
>> Signed-off-by: Steve Sistare <[email protected]>
>> Signed-off-by: Anthony Yznaga <[email protected]>
>> ---
>> include/uapi/asm-generic/mman-common.h | 3 +++
>> mm/madvise.c | 25 +++++++++++++++++++++++++
>> 2 files changed, 28 insertions(+)
>>
>> diff --git a/include/uapi/asm-generic/mman-common.h b/include/uapi/asm-generic/mman-common.h
>> index f94f65d429be..7c5f616b28f7 100644
>> --- a/include/uapi/asm-generic/mman-common.h
>> +++ b/include/uapi/asm-generic/mman-common.h
>> @@ -72,6 +72,9 @@
>> #define MADV_COLD 20 /* deactivate these pages */
>> #define MADV_PAGEOUT 21 /* reclaim these pages */
>>
>> +#define MADV_DOEXEC 22 /* do inherit across exec */
>> +#define MADV_DONTEXEC 23 /* don't inherit across exec */
>> +
>> /* compatibility flags */
>> #define MAP_FILE 0
>>
>> diff --git a/mm/madvise.c b/mm/madvise.c
>> index dd1d43cf026d..b447fa748649 100644
>> --- a/mm/madvise.c
>> +++ b/mm/madvise.c
>> @@ -103,6 +103,26 @@ static long madvise_behavior(struct vm_area_struct *vma,
>> case MADV_KEEPONFORK:
>> new_flags &= ~VM_WIPEONFORK;
>> break;
>> + case MADV_DOEXEC:
>
> For me MADV_KEEPONEXEC sounds better as it's symmetric to MADV_KEEPONFORK.

We chose MADV_DOEXEC and MADV_DONTEXEC to match the precedent set by:

#define MADV_DONTFORK 10 /* don't inherit across fork */


#define MADV_DOFORK 11 /* do inherit across fork */


I do like "keep" as a concise description of the operation. KEEPONFORK is not a perfect
analog because its opposite is wipe ...

#define MADV_WIPEONFORK 18 /* Zero memory on fork, child only */
#define MADV_KEEPONFORK 19 /* Undo MADV_WIPEONFORK */

... but if folks are ok with that then IMO these are all good choices:

MADV_KEEPONEXEC
MADV_DROPONEXEC

MADV_KEEPEXEC (shorter)
MADV_DROPEXEC

MADV_KEEP_EXEC (more legible, but no existing MADV names use 2nd underscores)
MADV_DROP_EXEC

Whatever folks like best.

- Steve

>> + /*
>> + * MADV_DOEXEC is only supported on private, non-executable,
>> + * non-stack anonymous memory and if the VM_EXEC_KEEP flag
>> + * is available.
>> + */
>> + if (!VM_EXEC_KEEP || vma->vm_file || vma->vm_flags & (VM_EXEC|VM_SHARED|VM_STACK)) {
>> + error = -EINVAL;
>> + goto out;
>> + }
>> + new_flags |= (new_flags & ~VM_MAYEXEC) | VM_EXEC_KEEP;
>> + break;
>> + case MADV_DONTEXEC:
>> + if (!VM_EXEC_KEEP) {
>> + error = -EINVAL;
>> + goto out;
>> + }
>> + if (new_flags & VM_EXEC_KEEP)
>> + new_flags |= (new_flags & ~VM_EXEC_KEEP) | VM_MAYEXEC;
>> + break;
>> case MADV_DONTDUMP:
>> new_flags |= VM_DONTDUMP;
>> break;
>> @@ -983,6 +1003,8 @@ static int madvise_inject_error(int behavior,
>> case MADV_SOFT_OFFLINE:
>> case MADV_HWPOISON:
>> #endif
>> + case MADV_DOEXEC:
>> + case MADV_DONTEXEC:
>> return true;
>>
>> default:
>> @@ -1037,6 +1059,9 @@ static int madvise_inject_error(int behavior,
>> * MADV_DONTDUMP - the application wants to prevent pages in the given range
>> * from being included in its core dump.
>> * MADV_DODUMP - cancel MADV_DONTDUMP: no longer exclude from core dump.
>> + * MADV_DOEXEC - On exec, preserve and duplicate this area in the new process
>> + * if the new process allows it.
>> + * MADV_DONTEXEC - Undo the effect of MADV_DOEXEC.
>> *
>> * return values:
>> * zero - success
>>
>