linux-user: Use memfd for open syscall emulation
PATCH 1/1 linux-user: Use memfd for open syscall emulation
For certain paths in /proc, the open syscall is intercepted and the returned file descriptor points to a temporary file with emulated contents.
If TMPDIR is not accessible or writable for the current user (for example in a read-only mounted chroot or container) tools such as ps from procps may fail unexpectedly. Trying to read one of these paths such as /proc/self/stat would return an error such as ENOENT or EROFS.
To relax the requirement on a writable TMPDIR, use memfd_create() instead to create an anonymous file and return its file descriptor.
--- linux-user/syscall.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-)
linux-user/syscall.c +9 -1
@@ -8265,9 +8265,11 @@ static int do_openat(CPUArchState *cpu_env, int dirfd, const char *pathname, int }if (fake_open->filename) { +int fd, r; + +#ifndef CONFIG_MEMFD const char *tmpdir;char filename[PATH_MAX ]; -int fd, r;/* create temporary file to map stat to */ tmpdir =getenv ("TMPDIR" );
@@ -8279,6 +8281,12 @@ static int do_openat(CPUArchState *cpu_env, int dirfd, const char *pathname, int return fd; }unlink (filename); +#else + fd =memfd_create ("qemu-open" ,0 ); +if (fd <0 ) { +return fd; + } +#endif if ((r = fake_open->fill (cpu_env, fd))) {int e = errno;
Signature · 1 line
2.25.1
Quoted 35 lines · Rainer Müller
For certain paths in /proc, the open syscall is intercepted and the returned file descriptor points to a temporary file with emulated contents.
If TMPDIR is not accessible or writable for the current user (for example in a read-only mounted chroot or container) tools such as ps from procps may fail unexpectedly. Trying to read one of these paths such as
/proc/self/statwould return an error such as ENOENT or EROFS.To relax the requirement on a writable TMPDIR, use
memfd_create()instead to create an anonymous file and return its file descriptor.Signed-off-by Rainer Müller <[email protected]>--- linux-user/syscall.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-)linux-user/syscall.c +3
-1@@ -8265,9 +8265,11 @@ static int do_openat(CPUArchState *cpu_env, int dirfd, const char *pathname, int }if (fake_open->filename) { +int fd, r; + +#ifndef CONFIG_MEMFD const char *tmpdir;char filename[PATH_MAX ]; -int fd, r;/* create temporary file to map stat to */ tmpdir =getenv ("TMPDIR" );
linux-user/syscall.c static int do_openat(CPUArchState *cpu_env, int dirfd, const char *pathname, int · line 8289
Quoted 9 lines
linux-user/syscall.c +6
-0@@ -8279,6 +8281,12 @@ static int do_openat(CPUArchState *cpu_env, int dirfd, const char *pathname, int return fd; }unlink (filename); +#else + fd =memfd_create ("qemu-open" ,0 ); +if (fd <0 ) { +return fd; + } +#endif
Even without CONFIG_MEMFD, we will have the memfd_create function available in util/. I think you should drop the ifdefs like so:
#include "qemu/memfd.h"
fd = memfd_create(...);
if (fd < 0) {
if (errno != ENOSYS) {
return fd;
}
// tmpdir fallback
}r~
PATCH v2 1/1 linux-user: Use memfd for open syscall emulation
In reply to Rainer MüllerFor certain paths in /proc, the open syscall is intercepted and the returned file descriptor points to a temporary file with emulated contents.
If TMPDIR is not accessible or writable for the current user (for example in a read-only mounted chroot or container) tools such as ps from procps may fail unexpectedly. Trying to read one of these paths such as /proc/self/stat would return an error such as ENOENT or EROFS.
To relax the requirement on a writable TMPDIR, use memfd_create() instead to create an anonymous file and return its file descriptor.
---
v2: no more #ifdefs, use stub from util/memfd.c with ENOSYS fallback,
tested with 'strace -e fault=memfd_create'
--- linux-user/syscall.c | 22 ++++++++++++++-------- 1 file changed, 14 insertions(+), 8 deletions(-)
linux-user/syscall.c +14 -8
@@ -8269,16 +8269,22 @@ static int do_openat(CPUArchState *cpu_env, int dirfd, const char *pathname, intchar filename[PATH_MAX ];int fd, r; -/* create temporary file to map stat to */ - tmpdir =getenv ("TMPDIR" ); -if (!tmpdir) - tmpdir ="/tmp" ; -snprintf (filename,sizeof (filename),"%s/qemu-open.XXXXXX" , tmpdir); - fd =mkstemp (filename); + fd =memfd_create ("qemu-open" ,0 );if (fd <0 ) { -return fd; +if (errno !=ENOSYS ) { +return fd; + } +/* create temporary file to map stat to */ + tmpdir =getenv ("TMPDIR" ); +if (!tmpdir) + tmpdir ="/tmp" ; +snprintf (filename,sizeof (filename),"%s/qemu-open.XXXXXX" , tmpdir); + fd =mkstemp (filename); +if (fd <0 ) { +return fd; + } +unlink (filename); } -unlink (filename);if ((r = fake_open->fill (cpu_env, fd))) {int e = errno;
Signature · 1 line
2.25.1
linux-user/syscall.c static int do_openat(CPUArchState *cpu_env, int dirfd, const char *pathname, int · line 8285
Quoted 9 lines · Rainer Müller
linux-user/syscall.c +9
-0+/* create temporary file to map stat to */ + tmpdir =getenv ("TMPDIR" ); +if (!tmpdir) + tmpdir ="/tmp" ; +snprintf (filename,sizeof (filename),"%s/qemu-open.XXXXXX" , tmpdir); + fd =mkstemp (filename); +if (fd <0 ) { +return fd; + }
We've been using g_file_open_tmp elsewhere; probably good to follow suit here.
r~
Quoted 12 lines · Richard Henderson
Quoted 10 lines · Rainer Müller
+ /* create temporary file to map stat to */ + tmpdir =
getenv("TMPDIR");
+ if (!tmpdir)
+ tmpdir = "/tmp";
+snprintf(filename, sizeof(filename), "%s/qemu-open.XXXXXX", tmpdir);
+ fd = mkstemp(filename);
+ if (fd < 0) {
+ return fd;
+ }
Quoted 2 lines
We've been using
g_file_open_tmpelsewhere; probably good to follow suit here.
That seemed reasonable at first, but with regards to error handling it gets a bit complicated.
The suggested g_file_open_tmp() would leave us with a GError only, but to return something meaningful to the caller we must set errno in this context. As far as I can see, there is no way to convert back to an errno from GError.
With g_file_open_tmp() we could always set the same generic errno, but that would hide the real cause completely. I debugged this problem with this message that was confusing, but at least it gave away a hint:
cat: can't open '/proc/self/stat': Read-only file system
The other option would be to g_assert_true(fd >= 0) and kill the process in case opening the temporary file failed. This also feels wrong, as the caller could still recover from this state and continue.
Rainer
Richard Henderson
Reviewed-byQuoted 24 lines · Rainer Müller
Quoted 12 lines · Richard Henderson
Quoted 10 lines · Rainer Müller
+ /* create temporary file to map stat to */ + tmpdir =
getenv("TMPDIR");
+ if (!tmpdir)
+ tmpdir = "/tmp";
+snprintf(filename, sizeof(filename), "%s/qemu-open.XXXXXX", tmpdir);
+ fd = mkstemp(filename);
+ if (fd < 0) {
+ return fd;
+ }Quoted 2 lines
We've been using
g_file_open_tmpelsewhere; probably good to follow suit here.That seemed reasonable at first, but with regards to error handling it gets a bit complicated.
The suggested
g_file_open_tmp()would leave us with a GError only, but to return something meaningful to the caller we must set errno in this context. As far as I can see, there is no way to convert back to an errno from GError.
Quoted 4 lines
With
g_file_open_tmp()we could always set the same generic errno, but that would hide the real cause completely. I debugged this problem with this message that was confusing, but at least it gave away a hint:
cat: can't open '/proc/self/stat': Read-only file system
That's a good reason. Let's leave this alone then.
r~