binfmt_elf_fdpic: auxvec updates.

PATCH In review Paul Mundt 3 participants 12 messages Jul 28, 2008 – Aug 4, 2008

Subsystems (3)

VFS SuperH exec and binfmt

lore

Share & tools

lore

Paul Mundt

author

2008-07-28 15:03 UTC #0 lore raw

cover binfmt_elf_fdpic: auxvec updates.

This is a tiny series of patches for bringing binfmt_elf_fdpic back in line with binfmt_elf in terms of auxiliary vector layout, particularly for the new entries that have popped up recently.

[PATCH 1/3] binfmt_elf_fdpic: Support auxvec base platform string. [PATCH 2/3] binfmt_elf_fdpic: Convert initial stack alignment to arch_align_stack(). [PATCH 3/3] binfmt_elf_fdpic: Wire up AT_EXECFD, AT_EXECFN, AT_SECURE.

These depend on the refactored NEW_AUX_ENT(), so against -next or -mm presently.

Signed-off-by Paul Mundt <[email protected]>

---

 fs/binfmt_elf_fdpic.c |   91 ++++++++++++++++++++++++++++++++++++--------------
 1 file changed, 66 insertions(+), 25 deletions(-)

Paul Mundt

author

2008-07-28 15:04 UTC #1 lore raw

PATCH 1/3 binfmt_elf_fdpic: Support auxvec base platform string.

Commit 483fad1c3fa1060d7e6710e84a065ad514571739 introduces AT_BASE_PLATFORM, but only implements it for binfmt_elf. Given that AT_VECTOR_SIZE_BASE is unconditionally enlarged for us, and it's only optionally added in for the platforms that set ELF_BASE_PLATFORM, wire it up for binfmt_elf_fdpic, too.

Signed-off-by Paul Mundt <[email protected]>

---

 fs/binfmt_elf_fdpic.c |   50 ++++++++++++++++++++++++++++++++++++++++++------
 1 files changed, 43 insertions(+), 7 deletions(-)

fs/binfmt_elf_fdpic.c +43 -7

index 80c1f95..0955d03 100644
@@ -455,6 +455,16 @@ error_kill:
 }
 
 /*****************************************************************************/
+
+#ifndef ELF_BASE_PLATFORM
+/*
+ * AT_BASE_PLATFORM indicates the "real" hardware/microarchitecture.
+ * If the arch defines ELF_BASE_PLATFORM (in asm/elf.h), the value
+ * will be copied to the user stack in the same manner as AT_PLATFORM.
+ */
+#define ELF_BASE_PLATFORM NULL
+#endif
+
 /*
  * present useful information to the program
  */

@@ -466,8 +476,8 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 	unsigned long sp, csp, nitems;
 	elf_caddr_t __user *argv, *envp;
 	size_t platform_len = 0, len;
-	char *k_platform;
-	char __user *u_platform, *p;
+	char *k_platform, *k_base_platform;
+	char __user *u_platform, *u_base_platform, *p;
 	long hwcap;
 	int loop;
 	int nr;	/* reset for each csp adjustment */

@@ -483,11 +493,14 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 		return -EFAULT;
 #endif
 
-	/* get hold of platform and hardware capabilities masks for the machine
-	 * we are running on.  In some cases (Sparc), this info is impossible
-	 * to get, in others (i386) it is merely difficult.
-	 */
 	hwcap = ELF_HWCAP;
+
+	/*
+	 * If this architecture has a platform capability string, copy it
+	 * to userspace.  In some cases (Sparc), this info is impossible
+	 * for userspace to get any other way, in others (i386) it is
+	 * merely difficult.
+	 */
 	k_platform = ELF_PLATFORM;
 	u_platform = NULL;

@@ -499,6 +512,21 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 			return -EFAULT;
 	}
 
+	/*
+	 * If this architecture has a "base" platform capability
+	 * string, copy it to userspace.
+	 */
+	k_base_platform = ELF_BASE_PLATFORM;
+	u_base_platform = NULL;
+
+	if (k_base_platform) {
+		platform_len = strlen(k_base_platform) + 1;
+		sp -= platform_len;
+		u_base_platform = (char __user *) sp;
+		if (__copy_to_user(u_base_platform, k_base_platform, platform_len) != 0)
+			return -EFAULT;
+	}
+
 #if defined(__i386__) && defined(CONFIG_SMP)
 	/* in some cases (e.g. Hyper-Threading), we want to avoid L1 evictions
 	 * by the processes running on the same package. One thing we can do is

@@ -543,7 +571,8 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 	/* force 16 byte _final_ alignment here for generality */
 #define DLINFO_ITEMS 13
 
-	nitems = 1 + DLINFO_ITEMS + (k_platform ? 1 : 0) + AT_VECTOR_SIZE_ARCH;
+	nitems = 1 + DLINFO_ITEMS + (k_platform ? 1 : 0) + \
+		(k_base_platform ? 1 : 0) + AT_VECTOR_SIZE_ARCH;
 
 	csp = sp;
 	sp -= nitems * 2 * sizeof(unsigned long);

@@ -575,6 +604,13 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 			    (elf_addr_t) (unsigned long) u_platform);
 	}
 
+	if (k_base_platform) {
+		nr = 0;
+		csp -= 2 * sizeof(unsigned long);
+		NEW_AUX_ENT(AT_BASE_PLATFORM,
+			    (elf_addr_t) (unsigned long) u_base_platform);
+	}
+
 	nr = 0;
 	csp -= DLINFO_ITEMS * 2 * sizeof(unsigned long);
 	NEW_AUX_ENT(AT_HWCAP,	hwcap);

Paul Mundt

author

2008-07-28 15:05 UTC #2 lore raw

PATCH 2/3 binfmt_elf_fdpic: Convert initial stack alignment to arch_align_stack().

binfmt_elf_fdpic seems to have grabbed a hard-coded hack from an ancient version of binfmt_elf in order to try and fix up initial stack alignment on multi-threaded x86, which while in addition to being unused, was also pushed down beyond the first set of operations on the stack pointer, negating the entire purpose.

These days, we have an architecture independent arch_align_stack(), so we switch to using that instead. Move the initial alignment up before the initial stores while we're at it.

Signed-off-by Paul Mundt <[email protected]>

---

 fs/binfmt_elf_fdpic.c |   27 ++++++++++-----------------
 1 files changed, 10 insertions(+), 17 deletions(-)

fs/binfmt_elf_fdpic.c +10 ~~-17~~

index 0955d03..bc64f5f 100644
@@ -482,11 +482,18 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 	int loop;
 	int nr;	/* reset for each csp adjustment */
 
-	/* we're going to shovel a whole load of stuff onto the stack */
+	/*
+	 * we're going to shovel a whole load of stuff onto the stack
+	 *
+	 * In some cases (e.g. Hyper-Threading), we want to avoid L1
+	 * evictions by the processes running on the same package. One
+	 * thing we can do is to shuffle the initial stack for them, so
+	 * we give the architecture an opportunity to do so here.
+	 */
 #ifdef CONFIG_MMU
-	sp = bprm->p;
+	sp = arch_align_stack(bprm->p);
 #else
-	sp = mm->start_stack;
+	sp = arch_align_stack(mm->start_stack);
 
 	/* stack the program arguments and environment */
 	if (elf_fdpic_transfer_args_to_stack(bprm, &sp) < 0)

@@ -527,20 +534,6 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 			return -EFAULT;
 	}
 
-#if defined(__i386__) && defined(CONFIG_SMP)
-	/* in some cases (e.g. Hyper-Threading), we want to avoid L1 evictions
-	 * by the processes running on the same package. One thing we can do is
-	 * to shuffle the initial stack for them.
-	 *
-	 * the conditionals here are unneeded, but kept in to make the code
-	 * behaviour the same as pre change unless we have hyperthreaded
-	 * processors. This keeps Mr Marcelo Person happier but should be
-	 * removed for 2.5
-	 */
-	if (smp_num_siblings > 1)
-		sp = sp - ((current->pid % 64) << 7);
-#endif
-
 	sp &= ~7UL;
 
 	/* stack the load map(s) */

Paul Mundt

author

2008-07-28 15:05 UTC #3 lore raw

PATCH 3/3 binfmt_elf_fdpic: Wire up AT_EXECFD, AT_EXECFN, AT_SECURE.

These auxvec entries are the only ones left unhandled out of the current base implementation. This syncs up binfmt_elf_fdpic with linux/auxvec.h and current binfmt_elf.

Signed-off-by Paul Mundt <[email protected]>

---

 fs/binfmt_elf_fdpic.c |   14 +++++++++++++-
 1 files changed, 13 insertions(+), 1 deletions(-)

fs/binfmt_elf_fdpic.c +13 -1

index bc64f5f..f0dd83c 100644
@@ -25,6 +25,7 @@
 #include <linux/fcntl.h>
 #include <linux/slab.h>
 #include <linux/pagemap.h>
+#include <linux/security.h>
 #include <linux/highmem.h>
 #include <linux/highuid.h>
 #include <linux/personality.h>

@@ -562,11 +563,14 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 	}
 
 	/* force 16 byte _final_ alignment here for generality */
-#define DLINFO_ITEMS 13
+#define DLINFO_ITEMS 15
 
 	nitems = 1 + DLINFO_ITEMS + (k_platform ? 1 : 0) + \
 		(k_base_platform ? 1 : 0) + AT_VECTOR_SIZE_ARCH;
 
+	if (bprm->interp_flags & BINPRM_FLAGS_EXECFD)
+		nitems++;
+
 	csp = sp;
 	sp -= nitems * 2 * sizeof(unsigned long);
 	sp -= (bprm->envc + 1) * sizeof(char *);	/* envv[] */

@@ -604,6 +608,12 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 			    (elf_addr_t) (unsigned long) u_base_platform);
 	}
 
+	if (bprm->interp_flags & BINPRM_FLAGS_EXECFD) {
+		nr = 0;
+		csp -= 2 * sizeof(unsigned long);
+		NEW_AUX_ENT(AT_EXECFD, bprm->interp_data);
+	}
+
 	nr = 0;
 	csp -= DLINFO_ITEMS * 2 * sizeof(unsigned long);
 	NEW_AUX_ENT(AT_HWCAP,	hwcap);

@@ -619,6 +629,8 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 	NEW_AUX_ENT(AT_EUID,	(elf_addr_t) current->euid);
 	NEW_AUX_ENT(AT_GID,	(elf_addr_t) current->gid);
 	NEW_AUX_ENT(AT_EGID,	(elf_addr_t) current->egid);
+	NEW_AUX_ENT(AT_SECURE,	security_bprm_secureexec(bprm));
+	NEW_AUX_ENT(AT_EXECFN,	bprm->exec);
 
 #ifdef ARCH_DLINFO
 	nr = 0;

David Howells

Acked-by on v1 3/3

2008-08-01 14:04 UTC #4 lore raw

Re: [PATCH 3/3] binfmt_elf_fdpic: Wire up AT_EXECFD, AT_EXECFN, AT_SECURE.

Quoted 5 lines · Paul Mundt

These auxvec entries are the only ones left unhandled out of the current base implementation. This syncs up binfmt_elf_fdpic with linux/auxvec.h and current binfmt_elf.
Signed-off-by Paul Mundt <[email protected]>

Acked-by David Howells <[email protected]>

Though I'm not sure bprm->exec is any use in NOMMU circumstances:-/

David Howells

2008-08-01 14:01 UTC #5 lore raw

Re: [PATCH 2/3] binfmt_elf_fdpic: Convert initial stack alignment to arch_align_stack().

In reply to Paul Mundt

fs/binfmt_elf_fdpic.c static int create_elf_fdpic_tables(struct linux_binprm *bprm, · line 496

Quoted 11 lines · Paul Mundt

fs/binfmt_elf_fdpic.c +7 -2

+	 * In some cases (e.g. Hyper-Threading), we want to avoid L1
+	 * evictions by the processes running on the same package. One
+	 * thing we can do is to shuffle the initial stack for them, so
+	 * we give the architecture an opportunity to do so here.
+	 */
 #ifdef CONFIG_MMU
-	sp = bprm->p;
+	sp = arch_align_stack(bprm->p);
 #else
-	sp = mm->start_stack;
+	sp = arch_align_stack(mm->start_stack);

Ummm... You're calling arch_align_stack() under NOMMU... Is that really a good idea?

You can't necessarily move the stack pointer without exiting the allocated region or shrinking the amount of stack space the executable asked for. If you want to do this sort of thing, you need to tell the memory allocator what you're up to - or at the very least allocate some slack.

David

Paul Mundt

author

2008-08-01 21:44 UTC #6 lore raw

Re: [PATCH 2/3] binfmt_elf_fdpic: Convert initial stack alignment to arch_align_stack().

Quoted 17 lines · David Howells

fs/binfmt_elf_fdpic.c static int create_elf_fdpic_tables(struct linux_binprm *bprm, · line 496
Quoted 11 lines · Paul Mundt
fs/binfmt_elf_fdpic.c +7 -2
+	 * In some cases (e.g. Hyper-Threading), we want to avoid L1
+	 * evictions by the processes running on the same package. One
+	 * thing we can do is to shuffle the initial stack for them, so
+	 * we give the architecture an opportunity to do so here.
+	 */
 #ifdef CONFIG_MMU
-	sp = bprm->p;
+	sp = arch_align_stack(bprm->p);
 #else
-	sp = mm->start_stack;
+	sp = arch_align_stack(mm->start_stack);
Ummm... You're calling arch_align_stack() under NOMMU... Is that really a good idea?

Not particularly, no.

Quoted 5 lines

You can't necessarily move the stack pointer without exiting the allocated region or shrinking the amount of stack space the executable asked for. If you want to do this sort of thing, you need to tell the memory allocator what you're up to - or at the very least allocate some slack.

Yes, that's a good point, and one that probably ought to be documented alongside the initial alignment. I'll send an updated patch.

Mike Frysinger

2008-08-04 03:24 UTC #7 lore raw

Re: [PATCH 2/3] binfmt_elf_fdpic: Convert initial stack alignment to arch_align_stack().

Quoted 19 lines · Paul Mundt

Quoted 13 lines · David Howells
fs/binfmt_elf_fdpic.c static int create_elf_fdpic_tables(struct linux_binprm *bprm, · line 496
Quoted 11 lines · Paul Mundt
fs/binfmt_elf_fdpic.c +7 -2
+    * In some cases (e.g. Hyper-Threading), we want to avoid L1
+    * evictions by the processes running on the same package. One
+    * thing we can do is to shuffle the initial stack for them, so
+    * we give the architecture an opportunity to do so here.
+    */
 #ifdef CONFIG_MMU
-   sp = bprm->p;
+   sp = arch_align_stack(bprm->p);
 #else
-   sp = mm->start_stack;
+   sp = arch_align_stack(mm->start_stack);
Quoted 2 lines
Ummm... You're calling arch_align_stack() under NOMMU... Is that really a good idea?
Not particularly, no.

Quoted 7 lines

Quoted 4 lines
You can't necessarily move the stack pointer without exiting the allocated region or shrinking the amount of stack space the executable asked for. If you want to do this sort of thing, you need to tell the memory allocator what you're up to - or at the very least allocate some slack.
Yes, that's a good point, and one that probably ought to be documented alongside the initial alignment. I'll send an updated patch.

fs/binfmt_elf_fdpic.c: In function 'create_elf_fdpic_tables':
fs/binfmt_elf_fdpic.c:497: error: implicit declaration of function 'arch_align_stack'
make[1]: *** [fs/binfmt_elf_fdpic.o] Error 1
make: *** [fs] Error 2

Blackfin lacks a arch_align_stack(x) in asm-blackfin/system.h ... i imagine a stub will work for us like every other arch though

feel like doing that Bryan ? i'm guessing Paul doesnt want to ... ;) -mike

Paul Mundt

author

2008-08-04 04:00 UTC #8 lore raw

Re: [PATCH 2/3] binfmt_elf_fdpic: Convert initial stack alignment to arch_align_stack().

Quoted 34 lines · Mike Frysinger

Quoted 19 lines · Paul Mundt
Quoted 13 lines · David Howells
Review
Quoted 11 lines · Paul Mundt
(quoted diff) +7 -2
+    * In some cases (e.g. Hyper-Threading), we want to avoid L1
+    * evictions by the processes running on the same package. One
+    * thing we can do is to shuffle the initial stack for them, so
+    * we give the architecture an opportunity to do so here.
+    */
 #ifdef CONFIG_MMU
-   sp = bprm->p;
+   sp = arch_align_stack(bprm->p);
 #else
-   sp = mm->start_stack;
+   sp = arch_align_stack(mm->start_stack);
Quoted 2 lines
Ummm... You're calling arch_align_stack() under NOMMU... Is that really a good idea?
Not particularly, no.
Quoted 7 lines
Quoted 4 lines
You can't necessarily move the stack pointer without exiting the allocated region or shrinking the amount of stack space the executable asked for. If you want to do this sort of thing, you need to tell the memory allocator what you're up to - or at the very least allocate some slack.
Yes, that's a good point, and one that probably ought to be documented alongside the initial alignment. I'll send an updated patch.
fs/binfmt_elf_fdpic.c: In function 'create_elf_fdpic_tables':
fs/binfmt_elf_fdpic.c:497: error: implicit declaration of function 'arch_align_stack'
make[1]: *** [fs/binfmt_elf_fdpic.o] Error 1
make: *** [fs] Error 2

Quoted 4 lines

Blackfin lacks a arch_align_stack(x) in asm-blackfin/system.h ... i imagine a stub will work for us like every other arch though
feel like doing that Bryan ? i'm guessing Paul doesnt want to ... ;)

There's no need, as David pointed out, that was the wrong thing to do for the nommu case anyways. Here's an updated patch to replace the previous one:

Signed-off-by Paul Mundt <[email protected]>

---

fs/binfmt_elf_fdpic.c +9 ~~-16~~

index 80c1f95..10c6cb8 100644
@@ -472,9 +472,16 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 	int loop;
 	int nr;	/* reset for each csp adjustment */
 
-	/* we're going to shovel a whole load of stuff onto the stack */
 #ifdef CONFIG_MMU
-	sp = bprm->p;
+	/*
+	 * we're going to shovel a whole load of stuff onto the stack
+	 *
+	 * In some cases (e.g. Hyper-Threading), we want to avoid L1
+	 * evictions by the processes running on the same package. One
+	 * thing we can do is to shuffle the initial stack for them, so
+	 * we give the architecture an opportunity to do so here.
+	 */
+	sp = arch_align_stack(bprm->p);
 #else
 	sp = mm->start_stack;

@@ -499,20 +506,6 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 			return -EFAULT;
 	}
 
-#if defined(__i386__) && defined(CONFIG_SMP)
-	/* in some cases (e.g. Hyper-Threading), we want to avoid L1 evictions
-	 * by the processes running on the same package. One thing we can do is
-	 * to shuffle the initial stack for them.
-	 *
-	 * the conditionals here are unneeded, but kept in to make the code
-	 * behaviour the same as pre change unless we have hyperthreaded
-	 * processors. This keeps Mr Marcelo Person happier but should be
-	 * removed for 2.5
-	 */
-	if (smp_num_siblings > 1)
-		sp = sp - ((current->pid % 64) << 7);
-#endif
-
 	sp &= ~7UL;
 
 	/* stack the load map(s) */

David Howells

2008-08-04 10:08 UTC #9 lore raw

Re: [PATCH 2/3] binfmt_elf_fdpic: Convert initial stack alignment to arch_align_stack().

Quoted 3 lines · Paul Mundt

There's no need, as David pointed out, that was the wrong thing to do for the nommu case anyways. Here's an updated patch to replace the previous one:

That's better, however:

Review

Quoted 5 lines

(quoted diff) +2 -2

-	/* we're going to shovel a whole load of stuff onto the stack */
 #ifdef CONFIG_MMU
-	sp = bprm->p;
+	/*
+	 * we're going to shovel a whole load of stuff onto the stack

I think that comment shouldn't be moved inside the conditional section since it applies to NOMMU as well. Picky, I know, but... It might actually be better to attach it to the banner comment for the whole function.

David
---
[PATCH] binfmt_elf_fdpic: Convert initial stack alignment to arch_align_stack()

From: Paul Mundt <[email protected]>

These days, we have an architecture independent arch_align_stack(), so we switch to using that instead. Move the initial alignment up before the initial stores while we're at it.

Signed-off-by Paul Mundt <[email protected]>

Signed-off-by David Howells <[email protected]>

---

 fs/binfmt_elf_fdpic.c |   25 ++++++++-----------------
 1 files changed, 8 insertions(+), 17 deletions(-)

fs/binfmt_elf_fdpic.c +8 ~~-17~~

index 7c83605..f9d6eaa 100644
@@ -466,7 +466,8 @@ error_kill:
 #endif
 
 /*
- * present useful information to the program
+ * present useful information to the program by shovelling it onto the new
+ * process's stack
  */
 static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 				   struct mm_struct *mm,

@@ -482,9 +483,13 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 	int loop;
 	int nr;	/* reset for each csp adjustment */
 
-	/* we're going to shovel a whole load of stuff onto the stack */
 #ifdef CONFIG_MMU
-	sp = bprm->p;
+	/* In some cases (e.g. Hyper-Threading), we want to avoid L1 evictions
+	 * by the processes running on the same package. One thing we can do is
+	 * to shuffle the initial stack for them, so we give the architecture
+	 * an opportunity to do so here.
+	 */
+	sp = arch_align_stack(bprm->p);
 #else
 	sp = mm->start_stack;

@@ -527,20 +532,6 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 			return -EFAULT;
 	}
 
-#if defined(__i386__) && defined(CONFIG_SMP)
-	/* in some cases (e.g. Hyper-Threading), we want to avoid L1 evictions
-	 * by the processes running on the same package. One thing we can do is
-	 * to shuffle the initial stack for them.
-	 *
-	 * the conditionals here are unneeded, but kept in to make the code
-	 * behaviour the same as pre change unless we have hyperthreaded
-	 * processors. This keeps Mr Marcelo Person happier but should be
-	 * removed for 2.5
-	 */
-	if (smp_num_siblings > 1)
-		sp = sp - ((current->pid % 64) << 7);
-#endif
-
 	sp &= ~7UL;
 
 	/* stack the load map(s) */

David Howells

2008-08-01 13:57 UTC #10 lore raw

Re: [PATCH 1/3] binfmt_elf_fdpic: Support auxvec base platform string.

In reply to Paul Mundt

fs/binfmt_elf_fdpic.c static int create_elf_fdpic_tables(struct linux_binprm *bprm, · line 575

Quoted 2 lines · Paul Mundt

fs/binfmt_elf_fdpic.c +2 -0

+	nitems = 1 + DLINFO_ITEMS + (k_platform ? 1 : 0) + \
+		(k_base_platform ? 1 : 0) + AT_VECTOR_SIZE_ARCH;

Why the backslash?

David

Paul Mundt

author

2008-08-01 21:46 UTC #11 lore raw

Re: [PATCH 1/3] binfmt_elf_fdpic: Support auxvec base platform string.

Quoted 7 lines · David Howells

fs/binfmt_elf_fdpic.c static int create_elf_fdpic_tables(struct linux_binprm *bprm, · line 575
Quoted 2 lines · Paul Mundt
fs/binfmt_elf_fdpic.c +2 -0
+	nitems = 1 + DLINFO_ITEMS + (k_platform ? 1 : 0) + \
+		(k_base_platform ? 1 : 0) + AT_VECTOR_SIZE_ARCH;
Why the backslash?

Habitual behavioural patterns?

If you feel strongly about it it can just be edited out of the patch.