On 28/3/23 19:47, Richard Henderson wrote:
On 3/28/23 06:51, Philippe Mathieu-Daudé wrote:
Hi Richard,
On 27/3/23 23:18, Richard Henderson wrote:
Pass the address of the last byte of the image, rather than
the first address past the last byte. This avoids overflow
when the last page of the address space is involved.
Signed-off-by: Richard Henderson <richard.hender...@linaro.org>
---
linux-user/elfload.c | 24 ++++++++++++------------
linux-user/flatload.c | 2 +-
2 files changed, 13 insertions(+), 13 deletions(-)
index e99570ca18..5efec2630e 100644
--- a/linux-user/flatload.c
+++ b/linux-user/flatload.c
@@ -448,7 +448,7 @@ static int load_flat_file(struct linux_binprm *
bprm,
* Allocate the address space.
*/
probe_guest_base(bprm->filename, 0,
- text_len + data_len + extra + indx_len);
+ text_len + data_len + extra + indx_len - 1);
Per linux-user/user-internals.h:
/**
* probe_guest_base:
* @image_name: the executable being loaded
* @loaddr: the lowest fixed address in the executable
* @hiaddr: the highest fixed address in the executable
*
* Creates the initial guest address space in the host memory space.
*
* If @loaddr == 0, then no address in the executable is fixed,
* i.e. it is fully relocatable. In that case @hiaddr is the size
* of the executable.
*
* This function will not return if a valid value for guest_base
* cannot be chosen. On return, the executable loader can expect
*
* target_mmap(loaddr, hiaddr - loaddr, ...)
*
* to succeed.
*/
Since here @loaddr == 0, "@hiaddr is the size of the executable",
not "the first address past the last byte".
So we can not "Pass the address of the last byte of the image"
to this API. Maybe the API description is incorrect, in that
case your patch is right. Otherwise we might need to tune
probe_guest_base().
I'll update the documentation.
Then:
Reviewed-by: Philippe Mathieu-Daudé <phi...@linaro.org>
