The previous revision was nacked by Torsten, but compared to the alternatives
at hand I think we should test this approach. Ideally we want all the complexity
of live-patching in the live-patching code and not in the patch. The other option
is to accept v4 and document the limitation to patch writers of not patching
functions > 8 arguments or marking such functions as notrace or equivalent
Changelog v6:
1. Experimental changes -- need loads of testing
Based on the assumption that very far TOC and LR values
indicate the call happened through a stub and the
stub return works differently from a local call which
uses klp_return_helper.
2. This now runs the test case posted by Petr
http://marc.info/?l=linux-kernel&m=145745300216069&w=2
Changelog v5:
1. Removed the mini-stack frame created for klp_return_helper.
As a result of the mini-stack frame, function with > 8
arguments could not be patched
2. Removed camel casing in the comments
Changelog v4:
1. Renamed klp_matchaddr() to klp_get_ftrace_location()
and used it just to convert the function address.
2. Synced klp_write_module_reloc() with s390(); made it
inline, no error message, return -ENOSYS
3. Added an error message when including
powerpc/include/asm/livepatch.h without HAVE_LIVEPATCH
4. Update some comments.
Changelog v3:
1. Moved -ENOSYS to -EINVAL in klp_write_module_reloc
2. Moved klp_matchaddr to use ftrace_location_range
Changelog v2:
1. Implement review comments by Michael
2. The previous version compared _NIP from the
wrong location to check for whether we
are going to a patched location
This patch enables live patching for powerpc. The current patch
is applied on top of topic/mprofile-kernel at
https://git.kernel.org/cgit/linux/kernel/git/powerpc/linux.git/
This patch builds on top of ftrace with regs changes and the
-mprofile-kernel changes. It detects a change in NIP after
the klp subsystem has potentially changes the NIP as a result
of a livepatch. In that case it saves the TOC in the parents
stack and the offset of the return address from the TOC in
the reserved (CR+4) space. This hack allows us to provide the
complete frame of the calling function as is to the caller
without having to create a mini-frame
Upon return from the patched function, the TOC and correct
LR is restored.
I tested the sample in the livepatch and an additional sample
that patches int_to_scsilun. I'll post out that sample if there
is interest later. I also tested ftrace functionality on the
command line to check for breakage
Signed-off-by: Torsten Duwe <[email protected]>
Signed-off-by: Balbir Singh <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
---
arch/powerpc/Kconfig | 3 ++
arch/powerpc/include/asm/livepatch.h | 47 +++++++++++++++++++++++
arch/powerpc/kernel/Makefile | 1 +
arch/powerpc/kernel/entry_64.S | 72 ++++++++++++++++++++++++++++++++++++
arch/powerpc/kernel/livepatch.c | 29 +++++++++++++++
include/linux/ftrace.h | 1 +
include/linux/livepatch.h | 2 +
kernel/livepatch/core.c | 28 ++++++++++++--
kernel/trace/ftrace.c | 14 ++++++-
9 files changed, 193 insertions(+), 4 deletions(-)
diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index 91da283..926c0ea 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
@@ -159,6 +159,7 @@ config PPC
select ARCH_HAS_DEVMEM_IS_ALLOWED
select HAVE_ARCH_SECCOMP_FILTER
select ARCH_HAS_UBSAN_SANITIZE_ALL
+ select HAVE_LIVEPATCH if HAVE_DYNAMIC_FTRACE_WITH_REGS
config GENERIC_CSUM
def_bool CPU_LITTLE_ENDIAN
@@ -1110,3 +1111,5 @@ config PPC_LIB_RHEAP
bool
source "arch/powerpc/kvm/Kconfig"
+
+source "kernel/livepatch/Kconfig"
diff --git a/arch/powerpc/include/asm/livepatch.h b/arch/powerpc/include/asm/livepatch.h
new file mode 100644
index 0000000..b9856ce
--- /dev/null
+++ b/arch/powerpc/include/asm/livepatch.h
@@ -0,0 +1,47 @@
+/*
+ * livepatch.h - powerpc-specific Kernel Live Patching Core
+ *
+ * Copyright (C) 2015 SUSE
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef _ASM_POWERPC64_LIVEPATCH_H
+#define _ASM_POWERPC64_LIVEPATCH_H
+
+#include <linux/module.h>
+
+#ifdef CONFIG_LIVEPATCH
+
+static inline int klp_check_compiler_support(void)
+{
+ return 0;
+}
+
+static inline int klp_write_module_reloc(struct module *mod, unsigned long
+ type, unsigned long loc, unsigned long value)
+{
+ /* This requires infrastructure changes; we need the loadinfos. */
+ return -ENOSYS;
+}
+
+static inline void klp_arch_set_pc(struct pt_regs *regs, unsigned long ip)
+{
+ regs->nip = ip;
+}
+
+#else /* CONFIG_LIVEPATCH */
+#error Include linux/livepatch.h, not asm/livepatch.h
+#endif /* CONFIG_LIVEPATCH */
+
+#endif /* _ASM_POWERPC64_LIVEPATCH_H */
diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile
index 2da380f..b767e14 100644
--- a/arch/powerpc/kernel/Makefile
+++ b/arch/powerpc/kernel/Makefile
@@ -119,6 +119,7 @@ obj-$(CONFIG_DYNAMIC_FTRACE) += ftrace.o
obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o
obj-$(CONFIG_FTRACE_SYSCALLS) += ftrace.o
obj-$(CONFIG_TRACING) += trace_clock.o
+obj-$(CONFIG_LIVEPATCH) += livepatch.o
ifneq ($(CONFIG_PPC_INDIRECT_PIO),y)
obj-y += iomap.o
diff --git a/arch/powerpc/kernel/entry_64.S b/arch/powerpc/kernel/entry_64.S
index ec7f8aa..01f53d4 100644
--- a/arch/powerpc/kernel/entry_64.S
+++ b/arch/powerpc/kernel/entry_64.S
@@ -1224,6 +1224,9 @@ _GLOBAL(ftrace_caller)
addi r3,r3,function_trace_op@toc@l
ld r5,0(r3)
+#ifdef CONFIG_LIVEPATCH
+ mr r14,r7 /* remember old NIP */
+#endif
/* Calculate ip from nip-4 into r3 for call below */
subi r3, r7, MCOUNT_INSN_SIZE
@@ -1248,6 +1251,9 @@ ftrace_call:
/* Load ctr with the possibly modified NIP */
ld r3, _NIP(r1)
mtctr r3
+#ifdef CONFIG_LIVEPATCH
+ cmpd r14,r3 /* has NIP been altered? */
+#endif
/* Restore gprs */
REST_8GPRS(0,r1)
@@ -1265,6 +1271,51 @@ ftrace_call:
ld r0, LRSAVE(r1)
mtlr r0
+#ifdef CONFIG_LIVEPATCH
+ beq+ 4f /* likely(old_NIP == new_NIP) */
+ /*
+ * For a local call, restore this TOC after calling the patch function.
+ * For a global call, it does not matter what we restore here,
+ * since the global caller does its own restore right afterwards,
+ * anyway. Just insert a klp_return_helper frame in any case,
+ * so a patch function can always count on the changed stack offsets.
+ * The patch introduces a frame such that from the patched function
+ * we return back to klp_return helper. For ABI compliance r12,
+ * lr and LRSAVE(r1) contain the address of klp_return_helper.
+ * We loaded ctr with the address of the patched function earlier
+ * We use the reserved space in the parent stack frame at CR+4.
+ * This space is stored with the contents of LR-TOC, where LR
+ * refers to the address to return to.
+ * Why do we need this?
+ * After patching we need to return to a trampoline return function
+ * that guarantees that we restore the TOC and return to the correct
+ * caller back
+ */
+klp_return_prepare: /* For debugging only */
+ subf r0, r2, r0 /* Calculate offset from current TOC */
+ sradi. r11, r0, 32
+ cmpdi r11, 0
+ bgt- 6f
+
+ bl 5f
+5: mflr r12
+ addi r12, r12, (klp_return_helper + 4 - .)@l
+ std r12, LRSAVE(r1)
+ mtlr r12
+ mfctr r12 /* allow for TOC calculation in newfunc */
+ std r2, 24(r1) /* save TOC now, unconditionally. */
+ stw r0, 12(r1) /* Of the final LR and save it in CR+4 */
+ bctr
+
+ /* If LR and TOC are very far, the callee should be restroing r2 */
+ /* Because we got there through a stub via create_stub */
+6: ld r0, LRSAVE(r1)
+ mtlr r0
+ mfctr r12
+ bctr
+4:
+#endif
+
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
stdu r1, -112(r1)
.globl ftrace_graph_call
@@ -1281,6 +1332,27 @@ _GLOBAL(ftrace_graph_stub)
_GLOBAL(ftrace_stub)
blr
+#ifdef CONFIG_LIVEPATCH
+/*
+ * Helper function for local calls that are becoming global
+ * due to live patching.
+ * We can't simply patch the NOP after the original call,
+ * because, depending on the consistency model, some kernel
+ * threads may still have called the original, local function
+ * *without* saving their TOC in the respective stack frame slot,
+ * so the decision is made per-thread during function return by
+ * maybe inserting a klp_return_helper frame or not.
+*/
+klp_return_helper:
+ ld r2, 24(r1) /* restore TOC (saved by ftrace_caller) */
+ lwa r0, 12(r1) /* Load from CR+4, offset of LR w.r.t TOC */
+ add r0, r0, r2 /* Add the offset to current TOC */
+ std r0, LRSAVE(r1) /* save the real return address */
+ mtlr r0
+ blr
+#endif
+
+
#else
_GLOBAL_TOC(_mcount)
/* Taken from output of objdump from lib64/glibc */
diff --git a/arch/powerpc/kernel/livepatch.c b/arch/powerpc/kernel/livepatch.c
new file mode 100644
index 0000000..17f1a14
--- /dev/null
+++ b/arch/powerpc/kernel/livepatch.c
@@ -0,0 +1,29 @@
+/*
+ * livepatch.c - powerpc-specific Kernel Live Patching Core
+ *
+ * Copyright (C) 2015 SUSE
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, see <http://www.gnu.org/licenses/>.
+ */
+#include <linux/ftrace.h>
+#include <asm/livepatch.h>
+
+/*
+ * Live patch works only with -mprofile-kernel on PPC. In this case,
+ * the ftrace location is always within the first 16 bytes.
+ */
+unsigned long klp_get_ftrace_location(unsigned long faddr)
+{
+ return ftrace_location_range(faddr, faddr + 16);
+}
diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
index 81de712..3481a8e 100644
--- a/include/linux/ftrace.h
+++ b/include/linux/ftrace.h
@@ -455,6 +455,7 @@ int ftrace_update_record(struct dyn_ftrace *rec, int enable);
int ftrace_test_record(struct dyn_ftrace *rec, int enable);
void ftrace_run_stop_machine(int command);
unsigned long ftrace_location(unsigned long ip);
+unsigned long ftrace_location_range(unsigned long start, unsigned long end);
unsigned long ftrace_get_addr_new(struct dyn_ftrace *rec);
unsigned long ftrace_get_addr_curr(struct dyn_ftrace *rec);
diff --git a/include/linux/livepatch.h b/include/linux/livepatch.h
index a882865..25a267b 100644
--- a/include/linux/livepatch.h
+++ b/include/linux/livepatch.h
@@ -134,6 +134,8 @@ int klp_unregister_patch(struct klp_patch *);
int klp_enable_patch(struct klp_patch *);
int klp_disable_patch(struct klp_patch *);
+unsigned long klp_get_ftrace_location(unsigned long faddr);
+
#endif /* CONFIG_LIVEPATCH */
#endif /* _LINUX_LIVEPATCH_H_ */
diff --git a/kernel/livepatch/core.c b/kernel/livepatch/core.c
index bc2c85c..e2dbf01 100644
--- a/kernel/livepatch/core.c
+++ b/kernel/livepatch/core.c
@@ -298,22 +298,39 @@ unlock:
rcu_read_unlock();
}
+/*
+ * Convert a function address into the appropriate ftrace location.
+ *
+ * The given address is returned on most architectures. Live patching
+ * usually works only when the ftrace location is the first instruction
+ * in the function.
+ */
+unsigned long __weak klp_get_ftrace_location(unsigned long faddr)
+{
+ return faddr;
+}
+
static void klp_disable_func(struct klp_func *func)
{
struct klp_ops *ops;
+ unsigned long ftrace_loc;
if (WARN_ON(func->state != KLP_ENABLED))
return;
if (WARN_ON(!func->old_addr))
return;
+ ftrace_loc = klp_get_ftrace_location(func->old_addr);
+ if (WARN_ON(!ftrace_loc))
+ return;
+
ops = klp_find_ops(func->old_addr);
if (WARN_ON(!ops))
return;
if (list_is_singular(&ops->func_stack)) {
WARN_ON(unregister_ftrace_function(&ops->fops));
- WARN_ON(ftrace_set_filter_ip(&ops->fops, func->old_addr, 1, 0));
+ WARN_ON(ftrace_set_filter_ip(&ops->fops, ftrace_loc, 1, 0));
list_del_rcu(&func->stack_node);
list_del(&ops->node);
@@ -328,6 +345,7 @@ static void klp_disable_func(struct klp_func *func)
static int klp_enable_func(struct klp_func *func)
{
struct klp_ops *ops;
+ unsigned long ftrace_loc;
int ret;
if (WARN_ON(!func->old_addr))
@@ -336,6 +354,10 @@ static int klp_enable_func(struct klp_func *func)
if (WARN_ON(func->state != KLP_DISABLED))
return -EINVAL;
+ ftrace_loc = klp_get_ftrace_location(func->old_addr);
+ if (WARN_ON(!ftrace_loc))
+ return -EINVAL;
+
ops = klp_find_ops(func->old_addr);
if (!ops) {
ops = kzalloc(sizeof(*ops), GFP_KERNEL);
@@ -352,7 +374,7 @@ static int klp_enable_func(struct klp_func *func)
INIT_LIST_HEAD(&ops->func_stack);
list_add_rcu(&func->stack_node, &ops->func_stack);
- ret = ftrace_set_filter_ip(&ops->fops, func->old_addr, 0, 0);
+ ret = ftrace_set_filter_ip(&ops->fops, ftrace_loc, 0, 0);
if (ret) {
pr_err("failed to set ftrace filter for function '%s' (%d)\n",
func->old_name, ret);
@@ -363,7 +385,7 @@ static int klp_enable_func(struct klp_func *func)
if (ret) {
pr_err("failed to register ftrace handler for function '%s' (%d)\n",
func->old_name, ret);
- ftrace_set_filter_ip(&ops->fops, func->old_addr, 1, 0);
+ ftrace_set_filter_ip(&ops->fops, ftrace_loc, 1, 0);
goto err;
}
diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
index eca592f..e1b3f23 100644
--- a/kernel/trace/ftrace.c
+++ b/kernel/trace/ftrace.c
@@ -1533,7 +1533,19 @@ static int ftrace_cmp_recs(const void *a, const void *b)
return 0;
}
-static unsigned long ftrace_location_range(unsigned long start, unsigned long end)
+/**
+ * ftrace_location_range - return the first address of a traced location
+ * if it touches the given ip range
+ * @start: start of range to search.
+ * @end: end of range to search (inclusive). @end points to the last byte
+ * to check.
+ *
+ * Returns rec->ip if the related ftrace location is a least partly within
+ * the given address range. That is, the first address of the instruction
+ * that is either a NOP or call to the function tracer. It checks the ftrace
+ * internal tables to determine if the address belongs or not.
+ */
+unsigned long ftrace_location_range(unsigned long start, unsigned long end)
{
struct ftrace_page *pg;
struct dyn_ftrace *rec;
--
2.5.0
On Wed, Mar 09, 2016 at 05:59:40PM +1100, Balbir Singh wrote:
>
> The previous revision was nacked by Torsten, but compared to the alternatives
I nacked it because I was confident it couldn't work. Same goes
for this one, sorry. My good intention was to save us all some work.
> @@ -1265,6 +1271,51 @@ ftrace_call:
> ld r0, LRSAVE(r1)
> mtlr r0
>
> +#ifdef CONFIG_LIVEPATCH
> + beq+ 4f /* likely(old_NIP == new_NIP) */
> + /*
> + * For a local call, restore this TOC after calling the patch function.
This is the key issue.
Ftrace_caller can gather and save the current r2 contents, no problem;
but the point is, it needs to be restored *after* the replacement function.
I see 3 ways to accomplish this:
1st: make _every_ replacement function aware of this, and make it restore
the TOC manually just before each return statement.
2nd: provide a global hook to do the job, and use a stack frame to execute it.
3rd: have a global hook like solution 2, but let it have its own data
structure, I'd call it a "shadow stack", for the real return addresses.
See struct fgraph_cpu_data in kernel/trace/trace_functions_graph.c
Using heuristics to determine whether the call was local or global
makes me feel highly uncomfortable; one day it will break and
nobody will remember why.
Balbir, the problem with your patch is that it goes only half the way from
my solution 2 towards solution 1. When you call a helper function on return,
you need a place to store the real return address.
I'll try to demonstrate a solution 1 as well, but you'll probably won't like
that either...
Torsten
On Wed 2016-03-09 10:19:04, Torsten Duwe wrote:
> On Wed, Mar 09, 2016 at 05:59:40PM +1100, Balbir Singh wrote:
> >
> > The previous revision was nacked by Torsten, but compared to the alternatives
>
> I nacked it because I was confident it couldn't work. Same goes
> for this one, sorry. My good intention was to save us all some work.
>
> > @@ -1265,6 +1271,51 @@ ftrace_call:
> > ld r0, LRSAVE(r1)
> > mtlr r0
> >
> > +#ifdef CONFIG_LIVEPATCH
> > + beq+ 4f /* likely(old_NIP == new_NIP) */
> > + /*
> > + * For a local call, restore this TOC after calling the patch function.
>
> This is the key issue.
>
> Ftrace_caller can gather and save the current r2 contents, no problem;
> but the point is, it needs to be restored *after* the replacement function.
> I see 3 ways to accomplish this:
>
> 1st: make _every_ replacement function aware of this, and make it restore
> the TOC manually just before each return statement.
>
> 2nd: provide a global hook to do the job, and use a stack frame to execute it.
>
> 3rd: have a global hook like solution 2, but let it have its own data
> structure, I'd call it a "shadow stack", for the real return addresses.
> See struct fgraph_cpu_data in kernel/trace/trace_functions_graph.c
>
> Using heuristics to determine whether the call was local or global
> makes me feel highly uncomfortable; one day it will break and
> nobody will remember why.
>
> Balbir, the problem with your patch is that it goes only half the way from
> my solution 2 towards solution 1. When you call a helper function on return,
> you need a place to store the real return address.
>
> I'll try to demonstrate a solution 1 as well, but you'll probably won't like
> that either...
To be honest, I still do not have a good picture about all the
problems in my head. Anyway, I would really appreciate if we could
find a solution that would work transparently. I mean that adding
an extra hacks into selected functions in the patch might be quite
error prone and problems hard to debug. I think that we all want this
but I wanted to be sure :-)
BTW: I am getting close to send a patch with some basic livepatch
documentation. It might be used to document "temporary" limitations.
Best Regards,
Petr
On Wed, Mar 09, 2016 at 10:44:23AM +0100, Petr Mladek wrote:
> find a solution that would work transparently. I mean that adding
> an extra hacks into selected functions in the patch might be quite
> error prone and problems hard to debug. I think that we all want this
> but I wanted to be sure :-)
Full ACK. Again, the TOC restore needs to go _after_ the replacement function,
and the klp_return_helper works as transparently as possible, so this
was my first choice. Arguments on the stack? I thought we'll deal with them
once we get there (e.g. _really_ need to patch a varargs function or one
with a silly signature).
Torsten
On Wed, 9 Mar 2016, Torsten Duwe wrote:
> > find a solution that would work transparently. I mean that adding
> > an extra hacks into selected functions in the patch might be quite
> > error prone and problems hard to debug. I think that we all want this
> > but I wanted to be sure :-)
>
> Full ACK. Again, the TOC restore needs to go _after_ the replacement function,
> and the klp_return_helper works as transparently as possible, so this
> was my first choice. Arguments on the stack? I thought we'll deal with them
> once we get there (e.g. _really_ need to patch a varargs function or one
> with a silly signature).
Well, the problem is, once such need arises, it's too late already.
You need to be able to patch the kernels which are already out there,
running on machines potentially for ages once all of a sudden there is a
CVE for >8args / varargs function.
At that time, only starting to put support in the kernel for that is waaay
to late :)
Thanks,
--
Jiri Kosina
SUSE Labs
On Wed, 2016-03-09 at 11:03 +0100, Torsten Duwe wrote:
> On Wed, Mar 09, 2016 at 10:44:23AM +0100, Petr Mladek wrote:
> > find a solution that would work transparently. I mean that adding
> > an extra hacks into selected functions in the patch might be quite
> > error prone and problems hard to debug. I think that we all want this
> > but I wanted to be sure :-)
>
> Full ACK. Again, the TOC restore needs to go _after_ the replacement function,
> and the klp_return_helper works as transparently as possible, so this
> was my first choice. Arguments on the stack? I thought we'll deal with them
> once we get there (e.g. _really_ need to patch a varargs function or one
> with a silly signature).
I agree it's unlikely many people will want to patch varargs functions, or
functions with stupid numbers of parameters.
But at least with the current proposals, we have no way of preventing them from
doing so. Which means the first sign they'll get that it doesn't work is when
they've applied the patch and their production system goes down. And not even
when they insert the patch, only when the patched function is called, possibly
some time later.
Now perhaps in reality most people are only applying livepatches from their
distro, in which case the distro should have tested it. But I don't know for
sure.
Still I'm happy for the current solution to go in (klp_return_helper creating a
minimal frame).
I think we can probably come up with a fully robust solution. But not tonight,
and not this week :)
cheers
On Wed, Mar 09, 2016 at 11:13:05AM +0100, Jiri Kosina wrote:
> On Wed, 9 Mar 2016, Torsten Duwe wrote:
> > was my first choice. Arguments on the stack? I thought we'll deal with them
> > once we get there (e.g. _really_ need to patch a varargs function or one
> > with a silly signature).
>
> Well, the problem is, once such need arises, it's too late already.
No, not if it's documented.
> You need to be able to patch the kernels which are already out there,
> running on machines potentially for ages once all of a sudden there is a
> CVE for >8args / varargs function.
Then you'd need a solution like I sent out yesterday, with a pre-prologue
caller that pops the extra frame, so the replacement can be more straight-
forward. Or you can just deal with the shifted offsets in the replacement.
I'll try to demonstrate the alternative. That would then be required for
_all_ replacement functions. Or can the live patching framework differentiate
and tell ftrace_caller whether to place a stack frame or not?
Miroslav? Petr? Can we have 2 sorts of replacement functions?
Torsten
On Wed 2016-03-09 12:16:47, Torsten Duwe wrote:
> On Wed, Mar 09, 2016 at 11:13:05AM +0100, Jiri Kosina wrote:
> > On Wed, 9 Mar 2016, Torsten Duwe wrote:
> > > was my first choice. Arguments on the stack? I thought we'll deal with them
> > > once we get there (e.g. _really_ need to patch a varargs function or one
> > > with a silly signature).
> >
> > Well, the problem is, once such need arises, it's too late already.
>
> No, not if it's documented.
>
> > You need to be able to patch the kernels which are already out there,
> > running on machines potentially for ages once all of a sudden there is a
> > CVE for >8args / varargs function.
>
> Then you'd need a solution like I sent out yesterday, with a pre-prologue
> caller that pops the extra frame, so the replacement can be more straight-
> forward. Or you can just deal with the shifted offsets in the replacement.
>
> I'll try to demonstrate the alternative. That would then be required for
> _all_ replacement functions. Or can the live patching framework differentiate
> and tell ftrace_caller whether to place a stack frame or not?
>
> Miroslav? Petr? Can we have 2 sorts of replacement functions?
I personally prefer to keep most functions without any special hack,
especially when it is needed only for one architecture. If a hack is
needed for "corner cases" and it is documented then, IMHO, we could
live with it for some time. We test all patches anyway, so.
But I could not speak for the LivePatching maintainers whose are Josh
and Jiri.
Best Regards,
Petr
On 09/03/16 20:19, Torsten Duwe wrote:
> On Wed, Mar 09, 2016 at 05:59:40PM +1100, Balbir Singh wrote:
>> The previous revision was nacked by Torsten, but compared to the alternatives
> I nacked it because I was confident it couldn't work. Same goes
> for this one, sorry. My good intention was to save us all some work.
I don't doubt that. I added it to the changelog to keep the history.
I've been working with the constraints we have to get a solution that
does not put the burden on the patch writer. That is why this is marked
experimental as it needs a lot of testing. I think we should mark livepatching
on PPC as experimental to begin with
>> @@ -1265,6 +1271,51 @@ ftrace_call:
>> ld r0, LRSAVE(r1)
>> mtlr r0
>>
>> +#ifdef CONFIG_LIVEPATCH
>> + beq+ 4f /* likely(old_NIP == new_NIP) */
>> + /*
>> + * For a local call, restore this TOC after calling the patch function.
> This is the key issue.
>
> Ftrace_caller can gather and save the current r2 contents, no problem;
> but the point is, it needs to be restored *after* the replacement function.
> I see 3 ways to accomplish this:
>
> 1st: make _every_ replacement function aware of this, and make it restore
> the TOC manually just before each return statement.
>
Yes and I think -pg without -mprofile-kernel does a good job of doing it.
In my patch I try to detect a call via stub and one without. The one with the
stub will do the right thing (global calls). For local calls I have the store
in CR+4 hook.
> 2nd: provide a global hook to do the job, and use a stack frame to execute it.
>
> 3rd: have a global hook like solution 2, but let it have its own data
> structure, I'd call it a "shadow stack", for the real return addresses.
> See struct fgraph_cpu_data in kernel/trace/trace_functions_graph.c
We thought of a shadow stack as well, but the copying can be expensive. I;ve
not looked at trace_functions_graph.c in detail, will look
> Using heuristics to determine whether the call was local or global
> makes me feel highly uncomfortable; one day it will break and
> nobody will remember why.
It could break, but as with any code, the code is only as good as the
test cases it passes :) We can document our design in detail
>
> Balbir, the problem with your patch is that it goes only half the way from
> my solution 2 towards solution 1. When you call a helper function on return,
> you need a place to store the real return address.
>
> I'll try to demonstrate a solution 1 as well, but you'll probably won't like
> that either...
Sure, look forward to it. I am keen on getting live-patching working. I think
v4 with the documented limitation is fine - see Michael's email as well
> Torsten
>
Thanks for looking into this,
Balbir Singh.
On Wed, 9 Mar 2016, Balbir Singh wrote:
>
> The previous revision was nacked by Torsten, but compared to the alternatives
> at hand I think we should test this approach. Ideally we want all the complexity
> of live-patching in the live-patching code and not in the patch. The other option
> is to accept v4 and document the limitation to patch writers of not patching
> functions > 8 arguments or marking such functions as notrace or equivalent
So I tried to read all the relevant emails and I must admit I am quite
lost. Unfortunately I cannot help much with powerpc part as my knowledge
is close to zero, but from live patching perspective there are only two
sustainable solutions (in my opinion). First, make it work transparently
for a patch writer. So no inline asm in patched functions. Second, make it
impossible to patch such problematic functions and document it as a
limitation. Well, this would be sad for sure.
So I think we are on the same page. Hopefully.
One or two nits follow.
> static void klp_disable_func(struct klp_func *func)
> {
> struct klp_ops *ops;
> + unsigned long ftrace_loc;
>
> if (WARN_ON(func->state != KLP_ENABLED))
> return;
> if (WARN_ON(!func->old_addr))
> return;
>
> + ftrace_loc = klp_get_ftrace_location(func->old_addr);
> + if (WARN_ON(!ftrace_loc))
> + return;
> +
WARN_ON here in klp_disable_func() is reasonable, because we registered a
stub for the function successfully, so something really wrong must
happened...
> ops = klp_find_ops(func->old_addr);
> if (WARN_ON(!ops))
> return;
>
> if (list_is_singular(&ops->func_stack)) {
> WARN_ON(unregister_ftrace_function(&ops->fops));
> - WARN_ON(ftrace_set_filter_ip(&ops->fops, func->old_addr, 1, 0));
> + WARN_ON(ftrace_set_filter_ip(&ops->fops, ftrace_loc, 1, 0));
>
> list_del_rcu(&func->stack_node);
> list_del(&ops->node);
> @@ -328,6 +345,7 @@ static void klp_disable_func(struct klp_func *func)
> static int klp_enable_func(struct klp_func *func)
> {
> struct klp_ops *ops;
> + unsigned long ftrace_loc;
> int ret;
>
> if (WARN_ON(!func->old_addr))
> @@ -336,6 +354,10 @@ static int klp_enable_func(struct klp_func *func)
> if (WARN_ON(func->state != KLP_DISABLED))
> return -EINVAL;
>
> + ftrace_loc = klp_get_ftrace_location(func->old_addr);
> + if (WARN_ON(!ftrace_loc))
> + return -EINVAL;
> +
But here it might be too strong. I think simple
if (!ftrace_loc) {
pr_err("...");
return -EINVAL;
}
would be enough I guess.
> ops = klp_find_ops(func->old_addr);
> if (!ops) {
> ops = kzalloc(sizeof(*ops), GFP_KERNEL);
> @@ -352,7 +374,7 @@ static int klp_enable_func(struct klp_func *func)
> INIT_LIST_HEAD(&ops->func_stack);
> list_add_rcu(&func->stack_node, &ops->func_stack);
>
> - ret = ftrace_set_filter_ip(&ops->fops, func->old_addr, 0, 0);
> + ret = ftrace_set_filter_ip(&ops->fops, ftrace_loc, 0, 0);
> if (ret) {
> pr_err("failed to set ftrace filter for function '%s' (%d)\n",
> func->old_name, ret);
> @@ -363,7 +385,7 @@ static int klp_enable_func(struct klp_func *func)
> if (ret) {
> pr_err("failed to register ftrace handler for function '%s' (%d)\n",
> func->old_name, ret);
> - ftrace_set_filter_ip(&ops->fops, func->old_addr, 1, 0);
> + ftrace_set_filter_ip(&ops->fops, ftrace_loc, 1, 0);
> goto err;
> }
Thinking about it, we need ftrace_loc only in cases where we call
ftrace_set_filter_ip() right? So we can move klp_get_ftrace_location()
call to appropriate if branch both in klp_disable_func() and
klp_enable_func().
Thanks,
Miroslav
On Wed, Mar 09, 2016 at 05:59:40PM +1100, Balbir Singh wrote:
>
> Changelog v6:
> 1. Experimental changes -- need loads of testing
> Based on the assumption that very far TOC and LR values
> indicate the call happened through a stub and the
> stub return works differently from a local call which
> uses klp_return_helper.
Well, this is true, but the inverse is not.
Less than 2GiB between LR and TOC is a necessary but not a sufficient
condition for a local call.
I see the problem with your code that it hardly detects calls from one module
to another nearby. It will fail iff you mistake a global call to a module loaded
earlier to be local while patching AND you access global data / 64-bit constants
after that call before the TOC is sanitised again.
Torsten