linux-user/syscall.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
The guest binary and libraries are not always map with the
executable bit in the host process. The guest may read a
/proc/self/maps with no executable address range. The
perm fields should be based on the guest permission inside
Qemu.
Signed-off-by: Nicolas Surbayrole <nsurbayrole@quarkslab.com>
---
linux-user/syscall.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index 389ec09764..0bbb2ff9c7 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -7888,9 +7888,9 @@ static int open_self_maps(void *cpu_env, int fd)
count = dprintf(fd, TARGET_ABI_FMT_ptr "-" TARGET_ABI_FMT_ptr
" %c%c%c%c %08" PRIx64 " %s %"PRId64,
h2g(min), h2g(max - 1) + 1,
- e->is_read ? 'r' : '-',
- e->is_write ? 'w' : '-',
- e->is_exec ? 'x' : '-',
+ (flags & PAGE_READ) ? 'r' : '-',
+ (flags & PAGE_WRITE_ORG) ? 'w' : '-',
+ (flags & PAGE_EXEC) ? 'x' : '-',
e->is_priv ? 'p' : '-',
(uint64_t) e->offset, e->dev, e->inode);
if (path) {
--
2.30.1
On 3/8/21 1:19 AM, Nicolas Surbayrole wrote: > The guest binary and libraries are not always map with the > executable bit in the host process. The guest may read a > /proc/self/maps with no executable address range. The > perm fields should be based on the guest permission inside > Qemu. > > Signed-off-by: Nicolas Surbayrole<nsurbayrole@quarkslab.com> > --- Reviewed-by: Richard Henderson <richard.henderson@linaro.org> r~
Le 08/03/2021 à 10:19, Nicolas Surbayrole a écrit : > The guest binary and libraries are not always map with the > executable bit in the host process. The guest may read a > /proc/self/maps with no executable address range. The > perm fields should be based on the guest permission inside > Qemu. > > Signed-off-by: Nicolas Surbayrole <nsurbayrole@quarkslab.com> > --- > linux-user/syscall.c | 6 +++--- > 1 file changed, 3 insertions(+), 3 deletions(-) > > diff --git a/linux-user/syscall.c b/linux-user/syscall.c > index 389ec09764..0bbb2ff9c7 100644 > --- a/linux-user/syscall.c > +++ b/linux-user/syscall.c > @@ -7888,9 +7888,9 @@ static int open_self_maps(void *cpu_env, int fd) > count = dprintf(fd, TARGET_ABI_FMT_ptr "-" TARGET_ABI_FMT_ptr > " %c%c%c%c %08" PRIx64 " %s %"PRId64, > h2g(min), h2g(max - 1) + 1, > - e->is_read ? 'r' : '-', > - e->is_write ? 'w' : '-', > - e->is_exec ? 'x' : '-', > + (flags & PAGE_READ) ? 'r' : '-', > + (flags & PAGE_WRITE_ORG) ? 'w' : '-', > + (flags & PAGE_EXEC) ? 'x' : '-', > e->is_priv ? 'p' : '-', > (uint64_t) e->offset, e->dev, e->inode); > if (path) { > Applied to my linux-user-for-6.0 branch. Thanks, Laurent
Le 08/03/2021 à 10:19, Nicolas Surbayrole a écrit : > The guest binary and libraries are not always map with the > executable bit in the host process. The guest may read a > /proc/self/maps with no executable address range. The > perm fields should be based on the guest permission inside > Qemu. > > Signed-off-by: Nicolas Surbayrole <nsurbayrole@quarkslab.com> > --- > linux-user/syscall.c | 6 +++--- > 1 file changed, 3 insertions(+), 3 deletions(-) > > diff --git a/linux-user/syscall.c b/linux-user/syscall.c > index 389ec09764..0bbb2ff9c7 100644 > --- a/linux-user/syscall.c > +++ b/linux-user/syscall.c > @@ -7888,9 +7888,9 @@ static int open_self_maps(void *cpu_env, int fd) > count = dprintf(fd, TARGET_ABI_FMT_ptr "-" TARGET_ABI_FMT_ptr > " %c%c%c%c %08" PRIx64 " %s %"PRId64, > h2g(min), h2g(max - 1) + 1, > - e->is_read ? 'r' : '-', > - e->is_write ? 'w' : '-', > - e->is_exec ? 'x' : '-', > + (flags & PAGE_READ) ? 'r' : '-', > + (flags & PAGE_WRITE_ORG) ? 'w' : '-', > + (flags & PAGE_EXEC) ? 'x' : '-', > e->is_priv ? 'p' : '-', > (uint64_t) e->offset, e->dev, e->inode); > if (path) { > It looks good. Alex, you wrote this code, any comment? Should we move this directly in read_self_maps() to have the guest values in MapInfo? Thanks, Laurent
On 3/8/21 2:47 AM, Laurent Vivier wrote: > Should we move this directly in read_self_maps() to have the guest values in MapInfo? No, because we also need read_self_maps() to create the guest in the first place. r~
Laurent Vivier <laurent@vivier.eu> writes: > Le 08/03/2021 à 10:19, Nicolas Surbayrole a écrit : >> The guest binary and libraries are not always map with the >> executable bit in the host process. The guest may read a >> /proc/self/maps with no executable address range. The >> perm fields should be based on the guest permission inside >> Qemu. >> >> Signed-off-by: Nicolas Surbayrole <nsurbayrole@quarkslab.com> >> --- >> linux-user/syscall.c | 6 +++--- >> 1 file changed, 3 insertions(+), 3 deletions(-) >> >> diff --git a/linux-user/syscall.c b/linux-user/syscall.c >> index 389ec09764..0bbb2ff9c7 100644 >> --- a/linux-user/syscall.c >> +++ b/linux-user/syscall.c >> @@ -7888,9 +7888,9 @@ static int open_self_maps(void *cpu_env, int fd) >> count = dprintf(fd, TARGET_ABI_FMT_ptr "-" TARGET_ABI_FMT_ptr >> " %c%c%c%c %08" PRIx64 " %s %"PRId64, >> h2g(min), h2g(max - 1) + 1, >> - e->is_read ? 'r' : '-', >> - e->is_write ? 'w' : '-', >> - e->is_exec ? 'x' : '-', >> + (flags & PAGE_READ) ? 'r' : '-', >> + (flags & PAGE_WRITE_ORG) ? 'w' : '-', >> + (flags & PAGE_EXEC) ? 'x' : '-', >> e->is_priv ? 'p' : '-', >> (uint64_t) e->offset, e->dev, e->inode); >> if (path) { >> > > It looks good. > > Alex, you wrote this code, any comment? I just checked to see if we need a Fixes but it seems we've been ignoring flags for anything other than page_check_range since the original parsing code so I think we are good. Acked-by: Alex Bennée <alex.bennee@linaro.org> > Should we move this directly in read_self_maps() to have the guest > values in MapInfo? Nope for the reasons Richard said. read_self_maps() is a QEMU internal function which we use elsewhere and needs to know the "real" truth ;-) > > Thanks, > Laurent -- Alex Bennée
© 2016 - 2024 Red Hat, Inc.