Skip to contentSkip to conversation

remove binfmts.h from header exports

PATCH In review Olaf Hering 3 participants 7 messages Aug 7, 2007 – Aug 8, 2007 exec and binfmt

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 remove binfmts.h from header exports

remove linux/binfmts.h from make headers_install

A recent patch added PAGE_SIZE to the part outside of __KERNEL__. qemu and ia32el have their own define of MAX_ARG_PAGES. No package uses linux/binfmts.h, so it is safe to not provide it.

---
 include/linux/Kbuild    |    1 -
 include/linux/binfmts.h |    2 --
 2 files changed, 3 deletions(-)
include/linux/Kbuild +0 -1
@@ -173,7 +173,6 @@ unifdef-y += atm.h
 unifdef-y += atm_tcp.h
 unifdef-y += audit.h
 unifdef-y += auto_fs.h
-unifdef-y += binfmts.h
 unifdef-y += capability.h
 unifdef-y += capi.h
 unifdef-y += cciss_ioctl.h
include/linux/binfmts.h +0 -2
@@ -17,7 +17,6 @@ struct pt_regs;
 /* sizeof(linux_binprm->buf) */
 #define BINPRM_BUF_SIZE 128
 
-#ifdef __KERNEL__
 
 #define CORENAME_MAX_SIZE 128
 
@@ -99,5 +98,4 @@ extern void compute_creds(struct linux_b
 extern int do_coredump(long signr, int exit_code, struct pt_regs * regs);
 extern int set_binfmt(struct linux_binfmt *new);
 
-#endif /* __KERNEL__ */
 #endif /* _LINUX_BINFMTS_H */
#1 lore raw
Quoted 4 lines · Olaf Hering

remove linux/binfmts.h from make headers_install

A recent patch added PAGE_SIZE to the part outside of __KERNEL__. qemu and ia32el have their own define of MAX_ARG_PAGES.

Should they use kernel header instead?

Quoted 1 line

No package uses linux/binfmts.h, so it is safe to not provide it.

And? Does it contain stuff which is userspace visible? binfmts.h has at least CORENAME_MAX_SIZE and SUID_DUMP_*

"userspace doesn't use header" is something headers_install has never been about.

Quoted 26 lines
include/linux/Kbuild +0 -1
@@ -173,7 +173,6 @@ unifdef-y += atm.h
 unifdef-y += atm_tcp.h
 unifdef-y += audit.h
 unifdef-y += auto_fs.h
-unifdef-y += binfmts.h
 unifdef-y += capability.h
 unifdef-y += capi.h
 unifdef-y += cciss_ioctl.h
include/linux/binfmts.h +0 -2
@@ -17,7 +17,6 @@ struct pt_regs;
 /* sizeof(linux_binprm->buf) */
 #define BINPRM_BUF_SIZE 128
 
-#ifdef __KERNEL__
 
 #define CORENAME_MAX_SIZE 128
 
@@ -99,5 +98,4 @@ extern void compute_creds(struct linux_b
 extern int do_coredump(long signr, int exit_code, struct pt_regs * regs);
 extern int set_binfmt(struct linux_binfmt *new);
 
-#endif /* __KERNEL__ */
 #endif /* _LINUX_BINFMTS_H */
#2 lore raw
Quoted 7 lines · Alexey Dobriyan
Quoted 4 lines · Olaf Hering

remove linux/binfmts.h from make headers_install

A recent patch added PAGE_SIZE to the part outside of __KERNEL__. qemu and ia32el have their own define of MAX_ARG_PAGES.

Should they use kernel header instead?

No, because the header will disappear. And from my understanding, the args limit is now gone.

Quoted 4 lines
Quoted 1 line

No package uses linux/binfmts.h, so it is safe to not provide it.

And? Does it contain stuff which is userspace visible? binfmts.h has at least CORENAME_MAX_SIZE and SUID_DUMP_*

Yes, its inside __KERNEL__. Have you read that header already? We are talking about the part below.
What part is useable for an application?

....
#include <linux/capability.h>

struct pt_regs;

/*
* These are the maximum length and maximum number of strings passed to the
* execve() system call. MAX_ARG_STRLEN is essentially random but serves to
* prevent the kernel from being unduly impacted by misaddressed pointers.
* MAX_ARG_STRINGS is chosen to fit in a signed 32-bit integer. */
#define MAX_ARG_STRLEN (PAGE_SIZE * 32)
#define MAX_ARG_STRINGS 0x7FFFFFFF

/* sizeof(linux_binprm->buf) */
#define BINPRM_BUF_SIZE 128
....

Quoted 2 lines

"userspace doesn't use header" is something headers_install has never been about.

Thats true.
I remember someone even wrote something for Documentation/ a few days ago.

#3 lore raw
Quoted 5 lines · Olaf Hering
Quoted 3 lines · Alexey Dobriyan
Quoted 2 lines · Olaf Hering

remove linux/binfmts.h from make headers_install

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

A recent patch added PAGE_SIZE to the part outside of __KERNEL__. qemu and ia32el have their own define of MAX_ARG_PAGES.

Should they use kernel header instead?

No, because the header will disappear. And from my understanding, the args limit is now gone.

OK.

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

No package uses linux/binfmts.h, so it is safe to not provide it.

And? Does it contain stuff which is userspace visible? binfmts.h has at least CORENAME_MAX_SIZE and SUID_DUMP_*

Yes, its inside __KERNEL__.

Can't you accept for a second that some stuff under __KERNEL__ was put there by mistake?

Quoted 1 line

We are talking about the part below.

No, we are talking about whole header since you're going to unexport whole header.

Quoted 3 lines

What part is useable for an application?

From part you quoted nothing. Otherwise:

fd = open("/proc/sys/fs/suid_dumpable", 1);
snprintf(buf, sizeof(buf), "%d", SUID_DUMP_ROOT);
write(fd, buf, strlen(buf));

Quoted 15 lines

#include <linux/capability.h>

struct pt_regs;

/*
* These are the maximum length and maximum number of strings passed to the
* execve() system call. MAX_ARG_STRLEN is essentially random but serves to
* prevent the kernel from being unduly impacted by misaddressed pointers.
* MAX_ARG_STRINGS is chosen to fit in a signed 32-bit integer. */
#define MAX_ARG_STRLEN (PAGE_SIZE * 32)
#define MAX_ARG_STRINGS 0x7FFFFFFF

/* sizeof(linux_binprm->buf) */
#define BINPRM_BUF_SIZE 128

#4 lore raw
Quoted 7 lines · Alexey Dobriyan
Quoted 4 lines
Quoted 2 lines

And? Does it contain stuff which is userspace visible? binfmts.h has at least CORENAME_MAX_SIZE and SUID_DUMP_*

Yes, its inside __KERNEL__.

Can't you accept for a second that some stuff under __KERNEL__ was put there by mistake?

I wonder why the defines exist at all? My grep doesnt find SUID_DUMP_*. Looking at the commit message from d6e711448137ca3301512cec41a2c2ce852b3d0a

...

Quoted 4 lines
Quoted 2 lines
        if (current->euid == current->uid && current->egid == current->gid)
                current->mm->dumpable = 1;

Should this be SUID_DUMP_USER?

Actually the feedback I had from last time was that the SUID_ defines should go because its clearer to follow the numbers. They can go everywhere (and there are lots of places where dumpable is tested/used as a bool in untouched code)
...

Exporting a random array size is not very useful. The app and the kernel has to check the string length anyway.

So lets remove the header and the 3 unused defines.

#5 lore raw
In reply to Olaf Hering
Quoted 5 lines · Olaf Hering

remove linux/binfmts.h from make headers_install

A recent patch added PAGE_SIZE to the part outside of __KERNEL__. qemu and ia32el have their own define of MAX_ARG_PAGES. No package uses linux/binfmts.h, so it is safe to not provide it.

How is "all packages" defined?

Rene.

#6 lore raw
Quoted 3 lines · Rene Herman
Quoted 1 line

No package uses linux/binfmts.h, so it is safe to not provide it.

How is "all packages" defined?

3052 openSuSE 10.3 packages.