| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-Id: <1235521315-3424-5-git-send-email-tabbott@mit.edu>
Date: Tue, 24 Feb 2009 19:21:55 -0500
From: Tim Abbott <tabbott@....EDU>
To: linux-kernel@...r.kernel.org
Cc: Tim Abbott <tabbott@....edu>, Anders Kaseorg <andersk@....edu>,
Jeff Arnold <jbarnold@....edu>, Waseem Daher <wdaher@....edu>,
Denys Vlasenko <vda.linux@...glemail.com>,
Nikanth Karthikesan <knikanth@...e.de>,
Rusty Russell <rusty@...tcorp.com.au>,
Andi Kleen <andi@...stfloor.org>
Subject: [PATCH v2 4/4] Ksplice: Support updating x86-32 and x86-64
From: Jeff Arnold <jbarnold@....edu>
Ksplice makes it possible to apply patches to the kernel without rebooting.
Signed-off-by: Jeff Arnold <jbarnold@....edu>
Signed-off-by: Anders Kaseorg <andersk@....edu>
Signed-off-by: Tim Abbott <tabbott@....edu>
Tested-by: Waseem Daher <wdaher@....edu>
---
Documentation/ksplice.txt | 304 ++++
MAINTAINERS | 10 +
arch/Kconfig | 14 +
arch/x86/Kconfig | 1 +
arch/x86/kernel/ksplice-arch.c | 96 ++
include/linux/kernel.h | 1 +
include/linux/ksplice.h | 224 +++
kernel/Makefile | 3 +
kernel/ksplice.c | 2995 ++++++++++++++++++++++++++++++++++++++++
kernel/panic.c | 1 +
10 files changed, 3649 insertions(+), 0 deletions(-)
create mode 100644 Documentation/ksplice.txt
create mode 100644 arch/x86/kernel/ksplice-arch.c
create mode 100644 include/linux/ksplice.h
create mode 100644 kernel/ksplice.c
diff --git a/Documentation/ksplice.txt b/Documentation/ksplice.txt
new file mode 100644
index 0000000..648d015
--- /dev/null
+++ b/Documentation/ksplice.txt
@@ -0,0 +1,304 @@
+Ksplice
+-------
+
+CONTENTS:
+
+1. Concepts: updates, module changes, target modules
+2. What changes can Ksplice handle?
+3. Dependency model
+4. Locking model
+5. altinstructions, smplocks, and parainstructions
+6. sysfs interface
+7. debugfs interface
+8. Hooks for running custom code during the update process
+
+0. Design Description
+---------------------
+
+For a description of the Ksplice design, please see the Ksplice
+technical overview document: <http://www.ksplice.com/doc/ksplice.pdf>.
+For usage examples and the Ksplice man pages, please see
+<http://www.ksplice.com>.
+
+The document below assumes familiarity with the Ksplice design and
+describes notable implementation details and the interface between the
+Ksplice kernel component and the Ksplice user space component.
+
+1. Concepts: updates, module changes, target modules
+----------------------------------------------------
+
+A Ksplice update (struct update) contains one or more Ksplice module
+changes (struct ksplice_mod_change), one for each target kernel module
+that should be changed by the update. ksplice_mod_changes are grouped
+together into a Ksplice update in order to allow multiple modules to
+be changed atomically.
+
+The contents of a Ksplice module change are documented via kernel-doc
+in include/linux/ksplice.h. To construct a new Ksplice update to be
+performed atomically, one needs to:
+ 1. Populate the fields of one or more ksplice_mod_change structures.
+ 2. Call the Ksplice function init_ksplice_mod_change() on each
+ ksplice_mod_change to register them with the Ksplice kernel
+ component. When init_ksplice_mod_change() is called on a
+ ksplice_mod_change, that ksplice_mod_change will be associated
+ with the other ksplice_mod_changes that share the same Ksplice
+ identifier (KID) field.
+ 3. After all of the mod_changes intended for a particular Ksplice
+ update have been loaded, that update can be applied via the sysfs
+ interface (described in Section 7 below).
+
+Each ksplice_mod_change has a two modules associated with it: an
+old_code module and a new_code module.
+
+A ksplice_mod_change's "old_code" module contains materials needed
+only for preparing the update. Specifically, the old_code module
+contains a copy of the pre-patch version of each of the compilation
+units changed by the ksplice_mod_change. The old_code module can be
+unloaded after the update has been applied in order to save memory.
+
+The ksplice_mod_change's "new_code" module contains the new sections
+to be inserted by the update; it needs to remain loaded for as long as
+the update is applied.
+
+Here's an example:
+
+Let's say that the Ksplice user space component wants to update the
+core kernel and the isdn module. The user space component will select
+a KID for this update (let's say 123abc) and generate four modules:
+
+ksplice_123abc_vmlinux_new (the new_code module for the vmlinux module change)
+ksplice_123abc_vmlinux_old (the old_code module for the vmlinux module change)
+ksplice_123abc_isdn_new (the new_code module for the isdn module change)
+ksplice_123abc_isdn_old (the old_code module for the isdn module change)
+
+Once both of the vmlinux modules have been loaded, one of the modules
+calls init_ksplice_mod_change on a ksplice_mod_change corresponding to
+the desired vmlinux changes.
+
+Similarly, once both of the isdn modules have been loaded, one of the
+modules calls init_ksplice_mod_change on a ksplice_mod_change
+corresponding to the desired isdn changes.
+
+Once all modules are loaded (in this example, four modules), the
+update can be applied atomically using the Ksplice sysfs interface.
+
+2. What changes can Ksplice handle?
+-----------------------------------
+
+The Ksplice user space component takes a source code patch and uses it
+to construct appropriate ksplice_mod_changes for an update. Ksplice
+can handle source code patches that add new functions, modify the text
+or arguments of existing functions, delete functions, move functions
+between compilation units, change functions from local to global (or
+vice versa), add exported symbols, rename exported symbols, and delete
+exported symbols. Ksplice can handle patches that modify either C
+code or assembly code.
+
+As described in the Ksplice technical overview document, a programmer
+needs to write some new code in order for Ksplice to apply a patch
+that makes semantic changes to kernel data structures. Some other
+limitations also apply:
+
+Ksplice does not support changes to __init functions that been
+unloaded from kernel memory. Ksplice also does not support changes to
+functions in .exit.text sections since Ksplice currently requires that
+all Ksplice updates affecting a module be reversed before that module
+can be unloaded.
+
+The Ksplice user space implementation does not currently support
+changes to weak symbols and changes to global read-only data
+structures (changes to read-only data structures that are local to a
+compilation unit are fine).
+
+Exported symbols:
+
+Ksplice can handle arbitrary changes to exported symbols in the source
+code patch.
+
+Ksplice deletes exported symbols by looking up the relevant struct
+kernel_symbol in the kernel's exported symbol table and replacing the
+name field with a pointer to a string that begins with DISABLED.
+
+Ksplice adds new exported symbols through the same mechanism; the
+relevant new_code module will have a ksymtab entry containing a symbol
+with a name beginning with DISABLED, and Ksplice will replace that
+with the name of the symbol to be exported when the update is
+atomically applied.
+
+Because the struct kernel_symbol for a newly exported symbol is
+contained in the Ksplice new_Code module, if a module using one of the
+newly exported symbols is loaded, that module will correctly depend on
+the Ksplice new_code module that exported the symbol.
+
+3. Dependency model
+-------------------
+
+Because Ksplice resolves symbols used in the post code using Ksplice
+relocations, Ksplice must enforce additional dependencies. Ksplice
+uses the use_module function to directly add dependencies on all the
+modules that the post code references.
+
+4. Locking model
+----------------
+
+From a locking perspective, Ksplice treats applying or removing a
+Ksplice update as analogous to loading or unloading a new version of
+the kernel modules patched by the update. Ksplice uses module_mutex
+to protect against a variety of race conditions related to modules
+being loaded or unloaded while Ksplice is applying or reversing an
+update; this approach also protects against race conditions involving
+multiple Ksplice updates being loaded or unloaded simultaneously as
+well.
+
+5. altinstructions, smplocks, and parainstructions
+--------------------------------------------------
+
+There are currently several mechanisms through which the Linux kernel
+will modify executable code at runtime.
+
+These mechanisms sometimes overwrite the storage unit of a relocation,
+which would cause problems if not handled properly by Ksplice.
+
+Ksplice solves this problem by writing "canary" bytes (e.g.,
+0x77777777) in the storage unit of the relocation in user space.
+Ksplice then checks whether the canary has been overwritten before
+using a Ksplice relocation to detect symbol values or to write a value
+to the storage unit of a Ksplice relocation.
+
+6. sysfs interface
+------------------
+
+Ksplice exports four sysfs files per Ksplice update in order to
+communicate with user space. For each update, these four files are
+located in a directory of the form /sys/kernel/ksplice/$kid, with $kid
+replaced by the KID of the Ksplice update.
+
+A. /sys/kernel/ksplice/$kid/stage (mode 0600)
+
+This file contains one of three strings:
+preparing: Indicates that this update has not yet been applied
+applied: Indicates that this update has been applied and has not been reversed
+reversed: Indicates that this update has been reversed
+
+When the stage is "preparing", the superuser can write "applied" to
+the stage file in order to instruct Ksplice to apply the update. When
+the stage is "applied", the superuser can write "reversed" to the
+stage file in order to instruct Ksplice to reverse the update.
+Afterwards, the superuser can write "cleanup" to the stage file in
+order to instruct Ksplice to clean up the debugging information and
+sysfs directory associated with the reversed update. Once an update
+is reversed, it cannot be re-applied without first cleaning up the
+update.
+
+B. /sys/kernel/ksplice/$kid/debug (mode 0600)
+
+The file contains a single number: 1 if debugging is enabled for this
+Ksplice update and 0 otherwise.
+
+The superuser can write a new value to this file to enable or disable
+debugging.
+
+C. /sys/kernel/ksplice/$kid/partial (mode 0600)
+
+The file contains a single number: 1 if the update should be applied
+even if some of the target modules are not loaded and 0 otherwise.
+
+D. /sys/kernel/ksplice/$kid/abort_cause (mode 0400)
+
+This file contains a value indicating either 1) that Ksplice
+successfully completed the most recently requested stage transition or
+2) why Ksplice aborted the most recently requested stage transition.
+
+Each abort_code string is described below, along with the stage
+transitions that might potentially trigger each possible abort code.
+The stage transitions are abbreviated as follows: preparing->applied
+(P->A), applied->reversed (A->R).
+
+ok (P->A, A->R): The most recent stage transition succeeded.
+
+no_match (P->A): Ksplice aborted the update because Ksplice was unable
+to match the old_code module's object code against the running
+kernel's object code.
+
+failed_to_find (P->A): Ksplice aborted the update because Ksplice was
+unable to resolve some of the symbols used in the update.
+
+missing_export (P->A): Ksplice aborted the update because the symbols
+exported by the kernel did not match Ksplice's expectations.
+
+already_reversed (P->A): Ksplice aborted the update because once an
+update has been reversed, it cannot be applied again (without first
+being cleaned up and reinitialized).
+
+module_busy (A->R): Ksplice aborted the undo operation because the
+target Ksplice update is in use by another kernel module;
+specifically, either the target Ksplice update exports a symbol that
+is in use by another module or another Ksplice update depends on this
+Ksplice update.
+
+out_of_memory (P->A, A->R): Ksplice aborted the operation because a call to
+kmalloc or vmalloc failed.
+
+code_busy (P->A, A->R): Ksplice aborted the operation because Ksplice
+was unable to find a moment when one or more of the to-be-patched
+functions was not a thread's kernel stack.
+
+target_not_loaded (P->A): Ksplice aborted the update because one of
+the target modules is not loaded and the partial option
+(/sys/kernel/ksplice/$kid/partial) is not enabled.
+
+call_failed (P->A, A->R): One of the calls included as part of this
+update returned nonzero exit status.
+
+unexpected_running_task (P->A, A->R): Ksplice aborted the operation
+because Ksplice observed a running task during the kernel stack check,
+at a time when Ksplice expected all tasks to be stopped by
+stop_machine.
+
+unexpected (P->A, A->R): Ksplice aborted the operation because it
+encountered an unspecified internal error. This condition can only be
+caused by an invalid input to Ksplice or a bug in Ksplice.
+
+E. /sys/kernel/ksplice/$kid/conflicts (mode 0400)
+
+This file is empty until Ksplice aborts an operation because of a
+code_busy condition (see "abort_code" above). This conflicts file
+then contains information about the process(es) that caused the stack
+check failure.
+
+Specifically, each line of this file consists of three space-separated
+values, describing a single conflict:
+
+$program_name $program_pid $conflict_label
+
+$program_name is the name of the program with the conflict.
+$program_pid is the pid of the program with the conflict.
+$conflict_label is the Ksplice label of the function with the conflict.
+
+7. debugfs interface
+--------------------
+
+Ksplice exports a single file to debugfs for each Ksplice update. The
+file has a name of the form ksplice_KID, where KID is the unique
+identifier of the Ksplice update. It contains debugging information
+in a human-readable format.
+
+8. Hooks for running custom code during the update process
+----------------------------------------------------------
+
+Ksplice allows a programmer to write custom code to be called from
+within the kernel during the update process. The kernel component
+allows custom code to be executed at the following times:
+
+- pre_apply: Called before the update has been applied and before the
+machine has been stopped. Allowed to fail.
+- check_apply: Called before the update has been applied but after the
+machine has been stopped. Allowed to fail.
+- apply: Called when the update is definitely going to be applied and
+when the machine is stopped. Not allowed to fail.
+- post_apply: Called when the update has been applied and the machine
+is no longer stopped. Not allowed to fail.
+- fail_apply: Called when the update failed to apply and the machine
+is no longer stopped. Not allowed to fail.
+
+Ksplice also provides analagous xxx_reverse hooks.
diff --git a/MAINTAINERS b/MAINTAINERS
index db65b4e..6b52212 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -2636,6 +2636,16 @@ W: http://miguelojeda.es/auxdisplay.htm
W: http://jair.lab.fi.uva.es/~migojed/auxdisplay.htm
S: Maintained
+KSPLICE
+P: Jeff Arnold
+M: jbarnold@...lice.com
+P: Anders Kaseorg
+M: andersk@...lice.com
+P: Tim Abbott
+M: tabbott@...lice.com
+W: http://www.ksplice.com
+S: Supported
+
LAPB module
L: linux-x25@...r.kernel.org
S: Orphan
diff --git a/arch/Kconfig b/arch/Kconfig
index 550dab2..d84326b 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -43,6 +43,17 @@ config KPROBES
for kernel debugging, non-intrusive instrumentation and testing.
If in doubt, say "N".
+config KSPLICE
+ tristate "Ksplice rebootless kernel updates"
+ depends on KALLSYMS_ALL && MODULE_UNLOAD && SYSFS && \
+ FUNCTION_DATA_SECTIONS
+ depends on HAVE_KSPLICE
+ help
+ Say Y here if you want to be able to apply certain kinds of
+ patches to your running kernel, without rebooting.
+
+ If unsure, say N.
+
config HAVE_EFFICIENT_UNALIGNED_ACCESS
bool
help
@@ -75,6 +86,9 @@ config HAVE_IOREMAP_PROT
config HAVE_KPROBES
bool
+config HAVE_KSPLICE
+ def_bool n
+
config HAVE_KRETPROBES
bool
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index d3a991c..bc3e68a 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -26,6 +26,7 @@ config X86
select HAVE_OPROFILE
select HAVE_IOREMAP_PROT
select HAVE_KPROBES
+ select HAVE_KSPLICE
select ARCH_WANT_OPTIONAL_GPIOLIB
select ARCH_WANT_FRAME_POINTERS
select HAVE_FUNCTION_DATA_SECTIONS
diff --git a/arch/x86/kernel/ksplice-arch.c b/arch/x86/kernel/ksplice-arch.c
new file mode 100644
index 0000000..c60bf79
--- /dev/null
+++ b/arch/x86/kernel/ksplice-arch.c
@@ -0,0 +1,96 @@
+/* Copyright (C) 2007-2009 Ksplice, Inc.
+ * Authors: Jeff Arnold, Anders Kaseorg, Tim Abbott
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2.
+ *
+ * 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, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+#define KSPLICE_IP(x) ((x)->thread.ip)
+#define KSPLICE_SP(x) ((x)->thread.sp)
+
+static struct ksplice_symbol trampoline_symbol = {
+ .name = NULL,
+ .label = "<trampoline>",
+};
+
+static const struct ksplice_reloc_howto trampoline_howto = {
+ .type = KSPLICE_HOWTO_RELOC,
+ .pcrel = 1,
+ .size = 4,
+ .dst_mask = 0xffffffffL,
+ .rightshift = 0,
+ .signed_addend = 1,
+};
+
+static const struct ksplice_reloc trampoline_reloc = {
+ .symbol = &trampoline_symbol,
+ .insn_addend = -4,
+ .target_addend = 0,
+ .howto = &trampoline_howto,
+};
+
+static abort_t trampoline_target(struct ksplice_mod_change *change,
+ unsigned long addr, unsigned long *new_addr)
+{
+ abort_t ret;
+ unsigned char byte;
+
+ if (probe_kernel_read(&byte, (void *)addr, sizeof(byte)) == -EFAULT)
+ return NO_MATCH;
+
+ if (byte != 0xe9)
+ return NO_MATCH;
+
+ ret = read_reloc_value(change, &trampoline_reloc, addr + 1, new_addr);
+ if (ret != OK)
+ return ret;
+
+ *new_addr += addr + 1;
+ return OK;
+}
+
+static abort_t prepare_trampoline(struct ksplice_mod_change *change,
+ struct ksplice_patch *p)
+{
+ p->size = 5;
+ ((unsigned char *)p->contents)[0] = 0xe9;
+ return write_reloc_value(change, &trampoline_reloc,
+ (unsigned long)p->contents + 1,
+ p->repladdr - (p->oldaddr + 1));
+}
+
+static abort_t handle_paravirt(struct ksplice_mod_change *change,
+ unsigned long pre_addr, unsigned long run_addr,
+ int *matched)
+{
+ unsigned char run[5], pre[5];
+ *matched = 0;
+
+ if (probe_kernel_read(&run, (void *)run_addr, sizeof(run)) == -EFAULT ||
+ probe_kernel_read(&pre, (void *)pre_addr, sizeof(pre)) == -EFAULT)
+ return OK;
+
+ if ((run[0] == 0xe8 && pre[0] == 0xe8) ||
+ (run[0] == 0xe9 && pre[0] == 0xe9))
+ if (run_addr + 1 + *(int32_t *)&run[1] ==
+ pre_addr + 1 + *(int32_t *)&pre[1])
+ *matched = 5;
+ return OK;
+}
+
+static bool valid_stack_ptr(const struct thread_info *tinfo, const void *p)
+{
+ return p > (const void *)tinfo
+ && p <= (const void *)tinfo + THREAD_SIZE - sizeof(long);
+}
+
diff --git a/include/linux/kernel.h b/include/linux/kernel.h
index 7fa3718..95990e1 100644
--- a/include/linux/kernel.h
+++ b/include/linux/kernel.h
@@ -307,6 +307,7 @@ extern enum system_states {
#define TAINT_OVERRIDDEN_ACPI_TABLE 8
#define TAINT_WARN 9
#define TAINT_CRAP 10
+#define TAINT_KSPLICE 11
extern void dump_stack(void) __cold;
diff --git a/include/linux/ksplice.h b/include/linux/ksplice.h
new file mode 100644
index 0000000..fbedd25
--- /dev/null
+++ b/include/linux/ksplice.h
@@ -0,0 +1,224 @@
+#include <linux/types.h>
+
+/**
+ * struct ksplice_symbol - Ksplice's analogue of an ELF symbol
+ * @name: The ELF name of the symbol
+ * @label: A unique Ksplice name for the symbol
+ * @candidate_vals: A linked list of possible values for the symbol, or NULL
+ * @value: The value of the symbol (valid when vals is NULL)
+ **/
+struct ksplice_symbol {
+ const char *name;
+ const char *label;
+/* private: */
+ struct list_head *candidate_vals;
+ unsigned long value;
+};
+
+/**
+ * struct ksplice_reloc - Ksplice's analogue of an ELF relocation
+ * @blank_addr: The address of the relocation's storage unit
+ * @symbol: The ksplice_symbol associated with this relocation
+ * @howto: The information regarding the relocation type
+ * @addend: The ELF addend of the relocation
+ **/
+struct ksplice_reloc {
+ unsigned long blank_addr;
+ struct ksplice_symbol *symbol;
+ const struct ksplice_reloc_howto *howto;
+ long insn_addend;
+ long target_addend;
+};
+
+enum ksplice_reloc_howto_type {
+ KSPLICE_HOWTO_RELOC,
+ KSPLICE_HOWTO_RELOC_PATCH,
+ KSPLICE_HOWTO_DATE,
+ KSPLICE_HOWTO_TIME,
+ KSPLICE_HOWTO_BUG,
+ KSPLICE_HOWTO_EXTABLE,
+};
+
+/**
+ * struct ksplice_reloc_howto - Ksplice's relocation type information
+ * @type: The type of the relocation
+ * @pcrel: Is the relocation PC relative?
+ * @size: The size, in bytes, of the item to be relocated
+ * @dst_mask: Bitmask for which parts of the instruction or data are
+ * replaced with the relocated value
+ * (based on dst_mask from GNU BFD's reloc_howto_struct)
+ * @rightshift: The value the final relocation is shifted right by;
+ * used to drop unwanted data from the relocation
+ * (based on rightshift from GNU BFD's reloc_howto_struct)
+ * @signed_addend: Should the addend be interpreted as a signed value?
+ **/
+struct ksplice_reloc_howto {
+ enum ksplice_reloc_howto_type type;
+ int pcrel;
+ int size;
+ long dst_mask;
+ unsigned int rightshift;
+ int signed_addend;
+};
+
+#if BITS_PER_LONG == 32
+#define KSPLICE_CANARY 0x77777777UL
+#elif BITS_PER_LONG == 64
+#define KSPLICE_CANARY 0x7777777777777777UL
+#endif /* BITS_PER_LONG */
+
+/**
+ * struct ksplice_section - Ksplice's analogue of an ELF section
+ * @symbol: The ksplice_symbol associated with this section
+ * @size: The length, in bytes, of this section
+ * @address: The address of the section
+ * @flags: Flags indicating the type of the section, whether or
+ * not it has been matched, etc.
+ **/
+struct ksplice_section {
+ struct ksplice_symbol *symbol;
+ unsigned long address;
+ unsigned long size;
+ unsigned int flags;
+ const unsigned char **match_map;
+};
+#define KSPLICE_SECTION_TEXT 0x00000001
+#define KSPLICE_SECTION_RODATA 0x00000002
+#define KSPLICE_SECTION_DATA 0x00000004
+#define KSPLICE_SECTION_STRING 0x00000008
+#define KSPLICE_SECTION_MATCHED 0x10000000
+
+#define MAX_TRAMPOLINE_SIZE 5
+
+enum ksplice_patch_type {
+ KSPLICE_PATCH_TEXT,
+ KSPLICE_PATCH_BUGLINE,
+ KSPLICE_PATCH_DATA,
+ KSPLICE_PATCH_EXPORT,
+};
+
+/**
+ * struct ksplice_patch - A replacement that Ksplice should perform
+ * @oldaddr: The address of the obsolete function or structure
+ * @repladdr: The address of the replacement function
+ * @type: The type of the ksplice patch
+ * @size: The size of the patch
+ * @contents: The bytes to be installed at oldaddr
+ * @vaddr The address of the page mapping used to write at oldaddr
+ * @saved: The bytes originally at oldaddr which were
+ * overwritten by the patch
+ **/
+struct ksplice_patch {
+ unsigned long oldaddr;
+ unsigned long repladdr;
+ enum ksplice_patch_type type;
+ unsigned int size;
+ void *contents;
+/* private: */
+ void *vaddr;
+ void *saved;
+};
+
+#ifdef __KERNEL__
+#include <linux/module.h>
+#include <linux/stringify.h>
+
+#define _PASTE(x, y) x##y
+#define PASTE(x, y) _PASTE(x, y)
+#define KSPLICE_UNIQ(s) PASTE(s##_, KSPLICE_MID)
+#define KSPLICE_KID_UNIQ(s) PASTE(s##_, KSPLICE_KID)
+
+/**
+ * struct ksplice_module_list_entry - A record of a ksplice_mod_change's target
+ * @target_mod_name: The name of the ksplice_mod_change's target module
+ * @new_code_mod_name: The name of the ksplice_mod_change's new_code module
+ * @applied: Whether the ksplice_mod_change was applied or not (this
+ * will be false for ksplice_mod_changes patching targets
+ * that are not loaded when the partial flag is set)
+ **/
+struct ksplice_module_list_entry {
+ const char *target_mod_name;
+ const char *new_code_mod_name;
+ const char *kid;
+ bool applied;
+/* private: */
+ struct list_head update_list; /* list head for this is per-update */
+ struct list_head list; /* list head for this is global */
+};
+
+/* List of all ksplice modules and the module they patch */
+extern struct list_head ksplice_modules;
+
+/* There are two actions, apply and reverse */
+#define KS_ACTIONS 2
+enum ksplice_action {
+ KS_APPLY,
+ KS_REVERSE,
+};
+
+/**
+ * struct ksplice_hooks - Hooks to be run during an action (apply or reverse)
+ * @pre: Runs before the action;
+ * may return nonzero to abort the action
+ * @check: Runs inside stop_machine before the action;
+ * may return nonzero to abort the action
+ * @intra: Runs inside stop_machine during the action
+ * @post: Runs after the action is successfully performed
+ * @fail: Runs if the action is aborted for any reason
+ */
+struct ksplice_hooks {
+ const typeof(int (*)(void)) *pre, *pre_end, *check, *check_end;
+ const typeof(void (*)(void)) *intra, *intra_end, *post, *post_end,
+ *fail, *fail_end;
+};
+
+/**
+ * struct ksplice_code - Ksplice metadata for an object
+ * @relocs: The Ksplice relocations for the object
+ * @symbols: The Ksplice symbols for the object
+ * @sections: The Ksplice sections for the object
+ **/
+struct ksplice_code {
+ struct ksplice_reloc *relocs, *relocs_end;
+ struct ksplice_section *sections, *sections_end;
+ struct ksplice_symbol *symbols, *symbols_end;
+};
+
+/**
+ * struct ksplice_mod_change - Data for one module modified by a Ksplice update
+ * @name: The name of the new_code module for the change
+ * @kid: The Ksplice unique identifier for the change
+ * @target_name: The name of the module modified by the change
+ * @new_code_mod: The new_code module for the change
+ * @old_code: The old code for run-pre matching
+ * @new_code: The new code to switch to
+ * @patches: The function replacements in the change
+ * @patches_end: The end pointer for patches array
+ * @hooks: Hooks to be run during apply and reverse
+ * @update: The atomic update the change is part of
+ * @target: The module modified by the change
+ * @safety_records: The ranges of addresses that must not be on a
+ * kernel stack for the patch to apply safely
+ **/
+struct ksplice_mod_change {
+ const char *name;
+ const char *kid;
+ const char *target_name;
+ struct module *new_code_mod;
+ struct ksplice_code old_code, new_code;
+ struct ksplice_patch *patches, *patches_end;
+ struct ksplice_hooks hooks[KS_ACTIONS];
+/* private: */
+ struct update *update;
+ struct module *target;
+ struct list_head temp_labelvals;
+ struct list_head safety_records;
+ struct list_head list;
+};
+
+
+int init_ksplice_mod_change(struct ksplice_mod_change *change);
+
+void cleanup_ksplice_mod_change(struct ksplice_mod_change *change);
+
+#endif /* __KERNEL__ */
diff --git a/kernel/Makefile b/kernel/Makefile
index 170a921..1a95df9 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -12,6 +12,8 @@ obj-y = sched.o fork.o exec_domain.o panic.o printk.o \
notifier.o ksysfs.o pm_qos_params.o sched_clock.o cred.o \
async.o
+CFLAGS_ksplice.o += -Iarch/$(SRCARCH)/kernel
+
ifdef CONFIG_FUNCTION_TRACER
# Do not trace debug files and internal ftrace files
CFLAGS_REMOVE_lockdep.o = -pg
@@ -71,6 +73,7 @@ obj-$(CONFIG_AUDIT) += audit.o auditfilter.o
obj-$(CONFIG_AUDITSYSCALL) += auditsc.o
obj-$(CONFIG_AUDIT_TREE) += audit_tree.o
obj-$(CONFIG_KPROBES) += kprobes.o
+obj-$(CONFIG_KSPLICE) += ksplice.o
obj-$(CONFIG_KGDB) += kgdb.o
obj-$(CONFIG_DETECT_SOFTLOCKUP) += softlockup.o
obj-$(CONFIG_GENERIC_HARDIRQS) += irq/
diff --git a/kernel/ksplice.c b/kernel/ksplice.c
new file mode 100644
index 0000000..f6528e0
--- /dev/null
+++ b/kernel/ksplice.c
@@ -0,0 +1,2995 @@
+/* Copyright (C) 2007-2009 Ksplice, Inc.
+ * Authors: Jeff Arnold, Anders Kaseorg, Tim Abbott
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2.
+ *
+ * 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, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+#include <linux/module.h>
+#include <linux/bug.h>
+#include <linux/ctype.h>
+#include <linux/debugfs.h>
+#include <linux/errno.h>
+#include <linux/kallsyms.h>
+#include <linux/kobject.h>
+#include <linux/kthread.h>
+#include <linux/pagemap.h>
+#include <linux/sched.h>
+#include <linux/sort.h>
+#include <linux/stop_machine.h>
+#include <linux/sysfs.h>
+#include <linux/time.h>
+#include <linux/uaccess.h>
+#include <linux/vmalloc.h>
+#include <linux/ksplice.h>
+
+enum stage {
+ STAGE_PREPARING, /* the update is not yet applied */
+ STAGE_APPLIED, /* the update is applied */
+ STAGE_REVERSED, /* the update has been applied and reversed */
+};
+
+/* parameter to modify run-pre matching */
+enum run_pre_mode {
+ RUN_PRE_INITIAL, /* dry run (only change temp_labelvals) */
+ RUN_PRE_DEBUG, /* dry run with byte-by-byte debugging */
+ RUN_PRE_FINAL, /* finalizes the matching */
+};
+
+enum { NOVAL, TEMP, VAL };
+
+typedef int __bitwise__ abort_t;
+
+#define OK ((__force abort_t) 0)
+#define NO_MATCH ((__force abort_t) 1)
+#define CODE_BUSY ((__force abort_t) 2)
+#define MODULE_BUSY ((__force abort_t) 3)
+#define OUT_OF_MEMORY ((__force abort_t) 4)
+#define FAILED_TO_FIND ((__force abort_t) 5)
+#define ALREADY_REVERSED ((__force abort_t) 6)
+#define MISSING_EXPORT ((__force abort_t) 7)
+#define UNEXPECTED_RUNNING_TASK ((__force abort_t) 8)
+#define UNEXPECTED ((__force abort_t) 9)
+#define TARGET_NOT_LOADED ((__force abort_t) 10)
+#define CALL_FAILED ((__force abort_t) 11)
+#define COLD_UPDATE_LOADED ((__force abort_t) 12)
+
+struct update {
+ const char *kid;
+ const char *name;
+ struct kobject kobj;
+ enum stage stage;
+ abort_t abort_cause;
+ int debug;
+#ifdef CONFIG_DEBUG_FS
+ struct debugfs_blob_wrapper debug_blob;
+ struct dentry *debugfs_dentry;
+#else /* !CONFIG_DEBUG_FS */
+ bool debug_continue_line;
+#endif /* CONFIG_DEBUG_FS */
+ bool partial; /* is it OK if some target mods aren't loaded */
+ struct list_head changes, /* changes for loaded target mods */
+ unused_changes; /* changes for non-loaded target mods */
+ struct list_head conflicts;
+ struct list_head list;
+ struct list_head ksplice_module_list;
+};
+
+/* a process conflicting with an update */
+struct conflict {
+ const char *process_name;
+ pid_t pid;
+ struct list_head stack;
+ struct list_head list;
+};
+
+/* an address on the stack of a conflict */
+struct conflict_addr {
+ unsigned long addr; /* the address on the stack */
+ bool has_conflict; /* does this address in particular conflict? */
+ const char *label; /* the label of the conflicting safety_record */
+ struct list_head list;
+};
+
+struct labelval {
+ struct list_head list;
+ struct ksplice_symbol *symbol;
+ struct list_head *saved_vals;
+};
+
+/* region to be checked for conflicts in the stack check */
+struct safety_record {
+ struct list_head list;
+ const char *label;
+ unsigned long addr; /* the address to be checked for conflicts
+ * (e.g. an obsolete function's starting addr)
+ */
+ unsigned long size; /* the size of the region to be checked */
+};
+
+/* possible value for a symbol */
+struct candidate_val {
+ struct list_head list;
+ unsigned long val;
+};
+
+/* private struct used by init_symbol_array */
+struct ksplice_lookup {
+/* input */
+ struct ksplice_mod_change *change;
+ struct ksplice_symbol **arr;
+ size_t size;
+/* output */
+ abort_t ret;
+};
+
+static LIST_HEAD(updates);
+LIST_HEAD(ksplice_modules);
+EXPORT_SYMBOL_GPL(ksplice_modules);
+static struct kobject *ksplice_kobj;
+
+static struct kobj_type update_ktype;
+
+static struct update *init_ksplice_update(const char *kid);
+static void cleanup_ksplice_update(struct update *update);
+static void maybe_cleanup_ksplice_update(struct update *update);
+static void add_to_update(struct ksplice_mod_change *change,
+ struct update *update);
+static int ksplice_sysfs_init(struct update *update);
+
+/* Preparing the relocations and patches for application */
+static abort_t apply_update(struct update *update);
+static abort_t reverse_update(struct update *update);
+static abort_t prepare_change(struct ksplice_mod_change *change);
+static abort_t finalize_change(struct ksplice_mod_change *change);
+static abort_t finalize_patches(struct ksplice_mod_change *change);
+static abort_t add_dependency_on_address(struct ksplice_mod_change *change,
+ unsigned long addr);
+static abort_t map_trampoline_pages(struct update *update);
+static void unmap_trampoline_pages(struct update *update);
+static void *map_writable(void *addr, size_t len);
+static abort_t apply_relocs(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *relocs,
+ const struct ksplice_reloc *relocs_end);
+static abort_t apply_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r);
+static abort_t apply_howto_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r);
+static abort_t apply_howto_date(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r);
+static abort_t read_reloc_value(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r,
+ unsigned long addr, unsigned long *valp);
+static abort_t write_reloc_value(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r,
+ unsigned long addr, unsigned long sym_addr);
+static abort_t create_module_list_entry(struct ksplice_mod_change *change,
+ bool to_be_applied);
+static void cleanup_module_list_entries(struct update *update);
+static void __attribute__((noreturn)) ksplice_deleted(void);
+
+/* run-pre matching */
+static abort_t match_change_sections(struct ksplice_mod_change *change,
+ bool consider_data_sections);
+static abort_t find_section(struct ksplice_mod_change *change,
+ struct ksplice_section *sect);
+static abort_t try_addr(struct ksplice_mod_change *change,
+ struct ksplice_section *sect,
+ unsigned long run_addr,
+ struct list_head *safety_records,
+ enum run_pre_mode mode);
+static abort_t run_pre_cmp(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect,
+ unsigned long run_addr,
+ struct list_head *safety_records,
+ enum run_pre_mode mode);
+static void print_bytes(struct ksplice_mod_change *change,
+ const unsigned char *run, int runc,
+ const unsigned char *pre, int prec);
+static const struct ksplice_reloc *
+init_reloc_search(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect);
+static const struct ksplice_reloc *find_reloc(const struct ksplice_reloc *start,
+ const struct ksplice_reloc *end,
+ unsigned long address,
+ unsigned long size);
+static abort_t lookup_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_reloc **fingerp,
+ unsigned long addr,
+ const struct ksplice_reloc **relocp);
+static abort_t handle_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect,
+ const struct ksplice_reloc *r,
+ unsigned long run_addr, enum run_pre_mode mode);
+static abort_t handle_howto_date(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect,
+ const struct ksplice_reloc *r,
+ unsigned long run_addr,
+ enum run_pre_mode mode);
+static abort_t handle_howto_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect,
+ const struct ksplice_reloc *r,
+ unsigned long run_addr,
+ enum run_pre_mode mode);
+#ifdef CONFIG_BUG
+static abort_t handle_bug(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r,
+ unsigned long run_addr);
+#endif /* CONFIG_BUG */
+static abort_t handle_extable(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r,
+ unsigned long run_addr);
+static struct ksplice_section *symbol_section(struct ksplice_mod_change *change,
+ const struct ksplice_symbol *sym);
+static int compare_section_labels(const void *va, const void *vb);
+static int symbol_section_bsearch_compare(const void *a, const void *b);
+static const struct ksplice_reloc *
+patch_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_patch *p);
+
+/* Computing possible addresses for symbols */
+static abort_t lookup_symbol(struct ksplice_mod_change *change,
+ const struct ksplice_symbol *ksym,
+ struct list_head *vals);
+static void cleanup_symbol_arrays(struct ksplice_mod_change *change);
+static abort_t init_symbol_arrays(struct ksplice_mod_change *change);
+static abort_t init_symbol_array(struct ksplice_mod_change *change,
+ struct ksplice_symbol *start,
+ struct ksplice_symbol *end);
+static abort_t uniquify_symbols(struct ksplice_mod_change *change);
+static abort_t add_matching_values(struct ksplice_lookup *lookup,
+ const char *sym_name, unsigned long sym_val);
+static bool add_export_values(const struct symsearch *syms,
+ struct module *owner,
+ unsigned int symnum, void *data);
+static int symbolp_bsearch_compare(const void *key, const void *elt);
+static int compare_symbolp_names(const void *a, const void *b);
+static int compare_symbolp_labels(const void *a, const void *b);
+static int add_kallsyms_values(void *data, const char *name,
+ struct module *owner, unsigned long val);
+static abort_t new_export_lookup(struct ksplice_mod_change *ichange,
+ const char *name, struct list_head *vals);
+
+/* Atomic update trampoline insertion and removal */
+static abort_t patch_action(struct update *update, enum ksplice_action action);
+static int __apply_patches(void *update);
+static int __reverse_patches(void *update);
+static abort_t check_each_task(struct update *update);
+static abort_t check_task(struct update *update,
+ const struct task_struct *t, bool rerun);
+static abort_t check_stack(struct update *update, struct conflict *conf,
+ const struct thread_info *tinfo,
+ const unsigned long *stack);
+static abort_t check_address(struct update *update,
+ struct conflict *conf, unsigned long addr);
+static abort_t check_record(struct conflict_addr *ca,
+ const struct safety_record *rec,
+ unsigned long addr);
+static bool is_stop_machine(const struct task_struct *t);
+static void cleanup_conflicts(struct update *update);
+static void print_conflicts(struct update *update);
+static void insert_trampoline(struct ksplice_patch *p);
+static abort_t verify_trampoline(struct ksplice_mod_change *change,
+ const struct ksplice_patch *p);
+static void remove_trampoline(const struct ksplice_patch *p);
+
+static abort_t create_labelval(struct ksplice_mod_change *change,
+ struct ksplice_symbol *ksym,
+ unsigned long val, int status);
+static abort_t create_safety_record(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect,
+ struct list_head *record_list,
+ unsigned long run_addr,
+ unsigned long run_size);
+static abort_t add_candidate_val(struct ksplice_mod_change *change,
+ struct list_head *vals, unsigned long val);
+static void release_vals(struct list_head *vals);
+static void set_temp_labelvals(struct ksplice_mod_change *change, int status);
+
+static int contains_canary(struct ksplice_mod_change *change,
+ unsigned long blank_addr,
+ const struct ksplice_reloc_howto *howto);
+static unsigned long follow_trampolines(struct ksplice_mod_change *change,
+ unsigned long addr);
+static bool patches_module(const struct module *a, const struct module *b);
+static bool strstarts(const char *str, const char *prefix);
+static bool singular(struct list_head *list);
+static void *bsearch(const void *key, const void *base, size_t n,
+ size_t size, int (*cmp)(const void *key, const void *elt));
+static int compare_relocs(const void *a, const void *b);
+static int reloc_bsearch_compare(const void *key, const void *elt);
+
+/* Debugging */
+static abort_t init_debug_buf(struct update *update);
+static void clear_debug_buf(struct update *update);
+static int __attribute__((format(printf, 2, 3)))
+_ksdebug(struct update *update, const char *fmt, ...);
+#define ksdebug(change, fmt, ...) \
+ _ksdebug(change->update, fmt, ## __VA_ARGS__)
+
+/* Architecture-specific functions defined in arch/ARCH/kernel/ksplice-arch.c */
+
+/* Prepare a trampoline for the given patch */
+static abort_t prepare_trampoline(struct ksplice_mod_change *change,
+ struct ksplice_patch *p);
+/* What address does the trampoline at addr jump to? */
+static abort_t trampoline_target(struct ksplice_mod_change *change,
+ unsigned long addr, unsigned long *new_addr);
+/* Hook to handle pc-relative jumps inserted by parainstructions */
+static abort_t handle_paravirt(struct ksplice_mod_change *change,
+ unsigned long pre, unsigned long run,
+ int *matched);
+/* Is address p on the stack of the given thread? */
+static bool valid_stack_ptr(const struct thread_info *tinfo, const void *p);
+
+#include "ksplice-arch.c"
+
+#define clear_list(head, type, member) \
+ do { \
+ struct list_head *_pos, *_n; \
+ list_for_each_safe(_pos, _n, head) { \
+ list_del(_pos); \
+ kfree(list_entry(_pos, type, member)); \
+ } \
+ } while (0)
+
+/**
+ * init_ksplice_mod_change() - Initializes a ksplice change
+ * @change: The change to be initialized. All of the public fields of the
+ * change and its associated data structures should be populated
+ * before this function is called. The values of the private
+ * fields will be ignored.
+ **/
+int init_ksplice_mod_change(struct ksplice_mod_change *change)
+{
+ struct update *update;
+ struct ksplice_patch *p;
+ struct ksplice_section *s;
+ int ret = 0;
+
+ INIT_LIST_HEAD(&change->temp_labelvals);
+ INIT_LIST_HEAD(&change->safety_records);
+
+ sort(change->old_code.relocs,
+ change->old_code.relocs_end - change->old_code.relocs,
+ sizeof(*change->old_code.relocs), compare_relocs, NULL);
+ sort(change->new_code.relocs,
+ change->new_code.relocs_end - change->new_code.relocs,
+ sizeof(*change->new_code.relocs), compare_relocs, NULL);
+ sort(change->old_code.sections,
+ change->old_code.sections_end - change->old_code.sections,
+ sizeof(*change->old_code.sections), compare_section_labels, NULL);
+
+ for (p = change->patches; p < change->patches_end; p++)
+ p->vaddr = NULL;
+ for (s = change->old_code.sections; s < change->old_code.sections_end;
+ s++)
+ s->match_map = NULL;
+ for (p = change->patches; p < change->patches_end; p++) {
+ const struct ksplice_reloc *r = patch_reloc(change, p);
+ if (r == NULL)
+ return -ENOENT;
+ if (p->type == KSPLICE_PATCH_DATA) {
+ s = symbol_section(change, r->symbol);
+ if (s == NULL)
+ return -ENOENT;
+ /* Ksplice creates KSPLICE_PATCH_DATA patches in order
+ * to modify rodata sections that have been explicitly
+ * marked for patching using the ksplice-patch.h macro
+ * ksplice_assume_rodata. Here we modify the section
+ * flags appropriately.
+ */
+ if (s->flags & KSPLICE_SECTION_DATA)
+ s->flags = (s->flags & ~KSPLICE_SECTION_DATA) |
+ KSPLICE_SECTION_RODATA;
+ }
+ }
+
+ mutex_lock(&module_mutex);
+ list_for_each_entry(update, &updates, list) {
+ if (strcmp(change->kid, update->kid) == 0) {
+ if (update->stage != STAGE_PREPARING) {
+ ret = -EPERM;
+ goto out;
+ }
+ add_to_update(change, update);
+ ret = 0;
+ goto out;
+ }
+ }
+ update = init_ksplice_update(change->kid);
+ if (update == NULL) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ ret = ksplice_sysfs_init(update);
+ if (ret != 0) {
+ cleanup_ksplice_update(update);
+ goto out;
+ }
+ add_to_update(change, update);
+out:
+ mutex_unlock(&module_mutex);
+ return ret;
+}
+EXPORT_SYMBOL_GPL(init_ksplice_mod_change);
+
+/**
+ * cleanup_ksplice_mod_change() - Cleans up a change if appropriate
+ * @change: The change to be cleaned up
+ *
+ * cleanup_ksplice_mod_change is currently called twice for each
+ * Ksplice update; once when the old_code module is unloaded, and once
+ * when the new_code module is unloaded. The extra call is used to
+ * avoid leaks if you unload the old_code without applying the update.
+ */
+void cleanup_ksplice_mod_change(struct ksplice_mod_change *change)
+{
+ if (change->update == NULL)
+ return;
+
+ mutex_lock(&module_mutex);
+ if (change->update->stage == STAGE_APPLIED) {
+ /* If the change wasn't actually applied (because we
+ * only applied this update to loaded modules and this
+ * target was not loaded), then unregister the change
+ * from the list of unused changes.
+ */
+ struct ksplice_mod_change *c;
+ bool found = false;
+
+ list_for_each_entry(c, &change->update->unused_changes, list) {
+ if (c == change)
+ found = true;
+ }
+ if (found)
+ list_del(&change->list);
+ mutex_unlock(&module_mutex);
+ return;
+ }
+ list_del(&change->list);
+ if (change->update->stage == STAGE_PREPARING)
+ maybe_cleanup_ksplice_update(change->update);
+ change->update = NULL;
+ mutex_unlock(&module_mutex);
+}
+EXPORT_SYMBOL_GPL(cleanup_ksplice_mod_change);
+
+static struct update *init_ksplice_update(const char *kid)
+{
+ struct update *update;
+ update = kcalloc(1, sizeof(struct update), GFP_KERNEL);
+ if (update == NULL)
+ return NULL;
+ update->name = kasprintf(GFP_KERNEL, "ksplice_%s", kid);
+ if (update->name == NULL) {
+ kfree(update);
+ return NULL;
+ }
+ update->kid = kstrdup(kid, GFP_KERNEL);
+ if (update->kid == NULL) {
+ kfree(update->name);
+ kfree(update);
+ return NULL;
+ }
+ if (try_module_get(THIS_MODULE) != 1) {
+ kfree(update->kid);
+ kfree(update->name);
+ kfree(update);
+ return NULL;
+ }
+ INIT_LIST_HEAD(&update->changes);
+ INIT_LIST_HEAD(&update->unused_changes);
+ INIT_LIST_HEAD(&update->ksplice_module_list);
+ if (init_debug_buf(update) != OK) {
+ module_put(THIS_MODULE);
+ kfree(update->kid);
+ kfree(update->name);
+ kfree(update);
+ return NULL;
+ }
+ list_add(&update->list, &updates);
+ update->stage = STAGE_PREPARING;
+ update->abort_cause = OK;
+ update->partial = 0;
+ INIT_LIST_HEAD(&update->conflicts);
+ return update;
+}
+
+static void cleanup_ksplice_update(struct update *update)
+{
+ list_del(&update->list);
+ cleanup_conflicts(update);
+ clear_debug_buf(update);
+ cleanup_module_list_entries(update);
+ kfree(update->kid);
+ kfree(update->name);
+ kfree(update);
+ module_put(THIS_MODULE);
+}
+
+/* Clean up the update if it no longer has any changes */
+static void maybe_cleanup_ksplice_update(struct update *update)
+{
+ if (list_empty(&update->changes) && list_empty(&update->unused_changes))
+ kobject_put(&update->kobj);
+}
+
+static void add_to_update(struct ksplice_mod_change *change,
+ struct update *update)
+{
+ change->update = update;
+ list_add(&change->list, &update->unused_changes);
+}
+
+static int ksplice_sysfs_init(struct update *update)
+{
+ int ret = 0;
+ memset(&update->kobj, 0, sizeof(update->kobj));
+ ret = kobject_init_and_add(&update->kobj, &update_ktype,
+ ksplice_kobj, "%s", update->kid);
+ if (ret != 0)
+ return ret;
+ kobject_uevent(&update->kobj, KOBJ_ADD);
+ return 0;
+}
+
+static abort_t apply_update(struct update *update)
+{
+ struct ksplice_mod_change *change, *n;
+ abort_t ret;
+ int retval;
+
+ list_for_each_entry(change, &update->changes, list) {
+ ret = create_module_list_entry(change, true);
+ if (ret != OK)
+ goto out;
+ }
+
+ list_for_each_entry_safe(change, n, &update->unused_changes, list) {
+ if (strcmp(change->target_name, "vmlinux") == 0) {
+ change->target = NULL;
+ } else if (change->target == NULL) {
+ change->target = find_module(change->target_name);
+ if (change->target == NULL ||
+ !module_is_live(change->target)) {
+ if (!update->partial) {
+ ret = TARGET_NOT_LOADED;
+ goto out;
+ }
+ ret = create_module_list_entry(change, false);
+ if (ret != OK)
+ goto out;
+ continue;
+ }
+ retval = use_module(change->new_code_mod,
+ change->target);
+ if (retval != 1) {
+ ret = UNEXPECTED;
+ goto out;
+ }
+ }
+ ret = create_module_list_entry(change, true);
+ if (ret != OK)
+ goto out;
+ list_del(&change->list);
+ list_add_tail(&change->list, &update->changes);
+
+ }
+
+ list_for_each_entry(change, &update->changes, list) {
+ const struct ksplice_section *sect;
+ for (sect = change->new_code.sections;
+ sect < change->new_code.sections_end; sect++) {
+ struct safety_record *rec = kmalloc(sizeof(*rec),
+ GFP_KERNEL);
+ if (rec == NULL) {
+ ret = OUT_OF_MEMORY;
+ goto out;
+ }
+ rec->addr = sect->address;
+ rec->size = sect->size;
+ rec->label = sect->symbol->label;
+ list_add(&rec->list, &change->safety_records);
+ }
+ }
+
+ list_for_each_entry(change, &update->changes, list) {
+ ret = init_symbol_arrays(change);
+ if (ret != OK) {
+ cleanup_symbol_arrays(change);
+ goto out;
+ }
+ ret = prepare_change(change);
+ cleanup_symbol_arrays(change);
+ if (ret != OK)
+ goto out;
+ }
+ ret = patch_action(update, KS_APPLY);
+out:
+ list_for_each_entry(change, &update->changes, list) {
+ struct ksplice_section *s;
+ if (update->stage == STAGE_PREPARING)
+ clear_list(&change->safety_records,
+ struct safety_record, list);
+ for (s = change->old_code.sections;
+ s < change->old_code.sections_end; s++) {
+ if (s->match_map != NULL) {
+ vfree(s->match_map);
+ s->match_map = NULL;
+ }
+ }
+ }
+ if (update->stage == STAGE_PREPARING)
+ cleanup_module_list_entries(update);
+
+ if (ret == OK)
+ printk(KERN_INFO "ksplice: Update %s applied successfully\n",
+ update->kid);
+ return ret;
+}
+
+static abort_t reverse_update(struct update *update)
+{
+ abort_t ret;
+ struct ksplice_mod_change *change;
+
+ clear_debug_buf(update);
+ ret = init_debug_buf(update);
+ if (ret != OK)
+ return ret;
+
+ _ksdebug(update, "Preparing to reverse %s\n", update->kid);
+
+ ret = patch_action(update, KS_REVERSE);
+ if (ret != OK)
+ return ret;
+
+ list_for_each_entry(change, &update->changes, list)
+ clear_list(&change->safety_records, struct safety_record, list);
+
+ printk(KERN_INFO "ksplice: Update %s reversed successfully\n",
+ update->kid);
+ return OK;
+}
+
+static int compare_symbolp_names(const void *a, const void *b)
+{
+ const struct ksplice_symbol *const *sympa = a, *const *sympb = b;
+ if ((*sympa)->name == NULL && (*sympb)->name == NULL)
+ return 0;
+ if ((*sympa)->name == NULL)
+ return -1;
+ if ((*sympb)->name == NULL)
+ return 1;
+ return strcmp((*sympa)->name, (*sympb)->name);
+}
+
+static int compare_symbolp_labels(const void *a, const void *b)
+{
+ const struct ksplice_symbol *const *sympa = a, *const *sympb = b;
+ return strcmp((*sympa)->label, (*sympb)->label);
+}
+
+static int symbolp_bsearch_compare(const void *key, const void *elt)
+{
+ const char *name = key;
+ const struct ksplice_symbol *const *symp = elt;
+ const struct ksplice_symbol *sym = *symp;
+ if (sym->name == NULL)
+ return 1;
+ return strcmp(name, sym->name);
+}
+
+static abort_t add_matching_values(struct ksplice_lookup *lookup,
+ const char *sym_name, unsigned long sym_val)
+{
+ struct ksplice_symbol **symp;
+ abort_t ret;
+
+ symp = bsearch(sym_name, lookup->arr, lookup->size,
+ sizeof(*lookup->arr), symbolp_bsearch_compare);
+ if (symp == NULL)
+ return OK;
+
+ while (symp > lookup->arr &&
+ symbolp_bsearch_compare(sym_name, symp - 1) == 0)
+ symp--;
+
+ for (; symp < lookup->arr + lookup->size; symp++) {
+ struct ksplice_symbol *sym = *symp;
+ if (sym->name == NULL || strcmp(sym_name, sym->name) != 0)
+ break;
+ ret = add_candidate_val(lookup->change,
+ sym->candidate_vals, sym_val);
+ if (ret != OK)
+ return ret;
+ }
+ return OK;
+}
+
+static int add_kallsyms_values(void *data, const char *name,
+ struct module *owner, unsigned long val)
+{
+ struct ksplice_lookup *lookup = data;
+ if (owner == lookup->change->new_code_mod ||
+ !patches_module(owner, lookup->change->target))
+ return (__force int)OK;
+ return (__force int)add_matching_values(lookup, name, val);
+}
+
+static bool add_export_values(const struct symsearch *syms,
+ struct module *owner,
+ unsigned int symnum, void *data)
+{
+ struct ksplice_lookup *lookup = data;
+ abort_t ret;
+
+ ret = add_matching_values(lookup, syms->start[symnum].name,
+ syms->start[symnum].value);
+ if (ret != OK) {
+ lookup->ret = ret;
+ return true;
+ }
+ return false;
+}
+
+static void cleanup_symbol_arrays(struct ksplice_mod_change *change)
+{
+ struct ksplice_symbol *sym;
+ for (sym = change->new_code.symbols; sym < change->new_code.symbols_end;
+ sym++) {
+ if (sym->candidate_vals != NULL) {
+ clear_list(sym->candidate_vals, struct candidate_val,
+ list);
+ kfree(sym->candidate_vals);
+ sym->candidate_vals = NULL;
+ }
+ }
+ for (sym = change->old_code.symbols; sym < change->old_code.symbols_end;
+ sym++) {
+ if (sym->candidate_vals != NULL) {
+ clear_list(sym->candidate_vals, struct candidate_val,
+ list);
+ kfree(sym->candidate_vals);
+ sym->candidate_vals = NULL;
+ }
+ }
+}
+
+/*
+ * The new_code and old_code modules each have their own independent
+ * ksplice_symbol structures. uniquify_symbols unifies these separate
+ * pieces of kernel symbol information by replacing all references to
+ * the old_code copy of symbols with references to the new_code copy.
+ */
+static abort_t uniquify_symbols(struct ksplice_mod_change *change)
+{
+ struct ksplice_reloc *r;
+ struct ksplice_section *s;
+ struct ksplice_symbol *sym, **sym_arr, **symp;
+ size_t size = change->new_code.symbols_end - change->new_code.symbols;
+
+ if (size == 0)
+ return OK;
+
+ sym_arr = vmalloc(sizeof(*sym_arr) * size);
+ if (sym_arr == NULL)
+ return OUT_OF_MEMORY;
+
+ for (symp = sym_arr, sym = change->new_code.symbols;
+ symp < sym_arr + size && sym < change->new_code.symbols_end;
+ sym++, symp++)
+ *symp = sym;
+
+ sort(sym_arr, size, sizeof(*sym_arr), compare_symbolp_labels, NULL);
+
+ for (r = change->old_code.relocs; r < change->old_code.relocs_end;
+ r++) {
+ symp = bsearch(&r->symbol, sym_arr, size, sizeof(*sym_arr),
+ compare_symbolp_labels);
+ if (symp != NULL) {
+ if ((*symp)->name == NULL)
+ (*symp)->name = r->symbol->name;
+ r->symbol = *symp;
+ }
+ }
+
+ for (s = change->old_code.sections; s < change->old_code.sections_end;
+ s++) {
+ symp = bsearch(&s->symbol, sym_arr, size, sizeof(*sym_arr),
+ compare_symbolp_labels);
+ if (symp != NULL) {
+ if ((*symp)->name == NULL)
+ (*symp)->name = s->symbol->name;
+ s->symbol = *symp;
+ }
+ }
+
+ vfree(sym_arr);
+ return OK;
+}
+
+/*
+ * Initialize the ksplice_symbol structures in the given array using
+ * the kallsyms and exported symbol tables.
+ */
+static abort_t init_symbol_array(struct ksplice_mod_change *change,
+ struct ksplice_symbol *start,
+ struct ksplice_symbol *end)
+{
+ struct ksplice_symbol *sym, **sym_arr, **symp;
+ struct ksplice_lookup lookup;
+ size_t size = end - start;
+ abort_t ret;
+
+ if (size == 0)
+ return OK;
+
+ for (sym = start; sym < end; sym++) {
+ if (strstarts(sym->label, "__ksymtab")) {
+ const struct kernel_symbol *ksym;
+ const char *colon = strchr(sym->label, ':');
+ const char *name = colon + 1;
+ if (colon == NULL)
+ continue;
+ ksym = find_symbol(name, NULL, NULL, true, false);
+ if (ksym == NULL) {
+ ksdebug(change, "Could not find kernel_symbol "
+ "structure for %s\n", name);
+ continue;
+ }
+ sym->value = (unsigned long)ksym;
+ sym->candidate_vals = NULL;
+ continue;
+ }
+
+ sym->candidate_vals = kmalloc(sizeof(*sym->candidate_vals),
+ GFP_KERNEL);
+ if (sym->candidate_vals == NULL)
+ return OUT_OF_MEMORY;
+ INIT_LIST_HEAD(sym->candidate_vals);
+ sym->value = 0;
+ }
+
+ sym_arr = vmalloc(sizeof(*sym_arr) * size);
+ if (sym_arr == NULL)
+ return OUT_OF_MEMORY;
+
+ for (symp = sym_arr, sym = start; symp < sym_arr + size && sym < end;
+ sym++, symp++)
+ *symp = sym;
+
+ sort(sym_arr, size, sizeof(*sym_arr), compare_symbolp_names, NULL);
+
+ lookup.change = change;
+ lookup.arr = sym_arr;
+ lookup.size = size;
+ lookup.ret = OK;
+
+ each_symbol(add_export_values, &lookup);
+ ret = lookup.ret;
+ if (ret == OK)
+ ret = (__force abort_t)
+ kallsyms_on_each_symbol(add_kallsyms_values, &lookup);
+ vfree(sym_arr);
+ return ret;
+}
+
+/* Prepare the change's ksplice_symbol structures for run-pre matching */
+static abort_t init_symbol_arrays(struct ksplice_mod_change *change)
+{
+ abort_t ret;
+
+ ret = uniquify_symbols(change);
+ if (ret != OK)
+ return ret;
+
+ ret = init_symbol_array(change, change->old_code.symbols,
+ change->old_code.symbols_end);
+ if (ret != OK)
+ return ret;
+
+ ret = init_symbol_array(change, change->new_code.symbols,
+ change->new_code.symbols_end);
+ if (ret != OK)
+ return ret;
+
+ return OK;
+}
+
+static abort_t prepare_change(struct ksplice_mod_change *change)
+{
+ abort_t ret;
+
+ ksdebug(change, "Preparing and checking %s\n", change->name);
+ ret = match_change_sections(change, false);
+ if (ret == NO_MATCH) {
+ /* It is possible that by using relocations from .data sections
+ * we can successfully run-pre match the rest of the sections.
+ * To avoid using any symbols obtained from .data sections
+ * (which may be unreliable) in the post code, we first prepare
+ * the post code and then try to run-pre match the remaining
+ * sections with the help of .data sections.
+ */
+ ksdebug(change, "Continuing without some sections; we might "
+ "find them later.\n");
+ ret = finalize_change(change);
+ if (ret != OK) {
+ ksdebug(change, "Aborted. Unable to continue without "
+ "the unmatched sections.\n");
+ return ret;
+ }
+
+ ksdebug(change, "run-pre: Considering .data sections to find "
+ "the unmatched sections\n");
+ ret = match_change_sections(change, true);
+ if (ret != OK)
+ return ret;
+
+ ksdebug(change, "run-pre: Found all previously unmatched "
+ "sections\n");
+ return OK;
+ } else if (ret != OK) {
+ return ret;
+ }
+
+ return finalize_change(change);
+}
+
+/*
+ * Finish preparing the change for insertion into the kernel.
+ * Afterwards, the replacement code should be ready to run and the
+ * ksplice_patches should all be ready for trampoline insertion.
+ */
+static abort_t finalize_change(struct ksplice_mod_change *change)
+{
+ abort_t ret;
+ ret = apply_relocs(change, change->new_code.relocs,
+ change->new_code.relocs_end);
+ if (ret != OK)
+ return ret;
+
+ ret = finalize_patches(change);
+ if (ret != OK)
+ return ret;
+
+ return OK;
+}
+
+static abort_t finalize_patches(struct ksplice_mod_change *change)
+{
+ struct ksplice_patch *p;
+ struct safety_record *rec;
+ abort_t ret;
+
+ for (p = change->patches; p < change->patches_end; p++) {
+ bool found = false;
+ list_for_each_entry(rec, &change->safety_records, list) {
+ if (rec->addr <= p->oldaddr &&
+ p->oldaddr < rec->addr + rec->size) {
+ found = true;
+ break;
+ }
+ }
+ if (!found && p->type != KSPLICE_PATCH_EXPORT) {
+ const struct ksplice_reloc *r = patch_reloc(change, p);
+ if (r == NULL) {
+ ksdebug(change, "A patch with no reloc at its "
+ "oldaddr has no safety record\n");
+ return NO_MATCH;
+ }
+ ksdebug(change, "No safety record for patch with"
+ "oldaddr %s+%lx\n", r->symbol->label,
+ r->target_addend);
+ return NO_MATCH;
+ }
+
+ if (p->type == KSPLICE_PATCH_TEXT) {
+ ret = prepare_trampoline(change, p);
+ if (ret != OK)
+ return ret;
+ }
+
+ if (found && rec->addr + rec->size < p->oldaddr + p->size) {
+ ksdebug(change, "Safety record %s is too short for "
+ "patch\n", rec->label);
+ return UNEXPECTED;
+ }
+
+ if (p->type == KSPLICE_PATCH_TEXT) {
+ if (p->repladdr == 0)
+ p->repladdr = (unsigned long)ksplice_deleted;
+ }
+ }
+ return OK;
+}
+
+/* noinline to prevent garbage on the stack from confusing check_stack */
+static noinline abort_t map_trampoline_pages(struct update *update)
+{
+ struct ksplice_mod_change *change;
+ list_for_each_entry(change, &update->changes, list) {
+ struct ksplice_patch *p;
+ for (p = change->patches; p < change->patches_end; p++) {
+ p->vaddr = map_writable((void *)p->oldaddr, p->size);
+ if (p->vaddr == NULL) {
+ ksdebug(change,
+ "Unable to map oldaddr read/write\n");
+ unmap_trampoline_pages(update);
+ return UNEXPECTED;
+ }
+ }
+ }
+ return OK;
+}
+
+static void unmap_trampoline_pages(struct update *update)
+{
+ struct ksplice_mod_change *change;
+ list_for_each_entry(change, &update->changes, list) {
+ struct ksplice_patch *p;
+ for (p = change->patches; p < change->patches_end; p++) {
+ vunmap((void *)((unsigned long)p->vaddr & PAGE_MASK));
+ p->vaddr = NULL;
+ }
+ }
+}
+
+/*
+ * map_writable creates a shadow page mapping of the range
+ * [addr, addr + len) so that we can write to code mapped read-only.
+ *
+ * It is similar to a generalized version of x86's text_poke. But
+ * because one cannot use vmalloc/vfree() inside stop_machine, we use
+ * map_writable to map the pages before stop_machine, then use the
+ * mapping inside stop_machine, and unmap the pages afterwards.
+ */
+static void *map_writable(void *addr, size_t len)
+{
+ void *vaddr;
+ int nr_pages = DIV_ROUND_UP(offset_in_page(addr) + len, PAGE_SIZE);
+ struct page **pages = kmalloc(nr_pages * sizeof(*pages), GFP_KERNEL);
+ void *page_addr = (void *)((unsigned long)addr & PAGE_MASK);
+ int i;
+
+ if (pages == NULL)
+ return NULL;
+
+ for (i = 0; i < nr_pages; i++) {
+ if (__module_address((unsigned long)page_addr) == NULL) {
+ pages[i] = virt_to_page(page_addr);
+ WARN_ON(!PageReserved(pages[i]));
+ } else {
+ pages[i] = vmalloc_to_page(addr);
+ }
+ if (pages[i] == NULL) {
+ kfree(pages);
+ return NULL;
+ }
+ page_addr += PAGE_SIZE;
+ }
+ vaddr = vmap(pages, nr_pages, VM_MAP, PAGE_KERNEL);
+ kfree(pages);
+ if (vaddr == NULL)
+ return NULL;
+ return vaddr + offset_in_page(addr);
+}
+
+/*
+ * Ksplice adds a dependency on any symbol address used to resolve
+ * relocations in the new_code module.
+ *
+ * Be careful to follow_trampolines so that we always depend on the
+ * latest version of the target function, since that's the code that
+ * will run if we call addr.
+ */
+static abort_t add_dependency_on_address(struct ksplice_mod_change *change,
+ unsigned long addr)
+{
+ struct ksplice_mod_change *c;
+ struct module *m =
+ __module_text_address(follow_trampolines(change, addr));
+ if (m == NULL)
+ return OK;
+ list_for_each_entry(c, &change->update->changes, list) {
+ if (m == c->new_code_mod)
+ return OK;
+ }
+ if (use_module(change->new_code_mod, m) != 1)
+ return MODULE_BUSY;
+ return OK;
+}
+
+static abort_t apply_relocs(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *relocs,
+ const struct ksplice_reloc *relocs_end)
+{
+ const struct ksplice_reloc *r;
+ for (r = relocs; r < relocs_end; r++) {
+ abort_t ret = apply_reloc(change, r);
+ if (ret != OK)
+ return ret;
+ }
+ return OK;
+}
+
+static abort_t apply_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r)
+{
+ switch (r->howto->type) {
+ case KSPLICE_HOWTO_RELOC:
+ case KSPLICE_HOWTO_RELOC_PATCH:
+ return apply_howto_reloc(change, r);
+ case KSPLICE_HOWTO_DATE:
+ case KSPLICE_HOWTO_TIME:
+ return apply_howto_date(change, r);
+ default:
+ ksdebug(change, "Unexpected howto type %d\n", r->howto->type);
+ return UNEXPECTED;
+ }
+}
+
+/*
+ * Applies a relocation. Aborts if the symbol referenced in it has
+ * not been uniquely resolved.
+ */
+static abort_t apply_howto_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r)
+{
+ abort_t ret;
+ int canary_ret;
+ unsigned long sym_addr;
+ LIST_HEAD(vals);
+
+ canary_ret = contains_canary(change, r->blank_addr, r->howto);
+ if (canary_ret < 0)
+ return UNEXPECTED;
+ if (canary_ret == 0) {
+ ksdebug(change, "reloc: skipped %lx to %s+%lx (altinstr)\n",
+ r->blank_addr, r->symbol->label, r->target_addend);
+ return OK;
+ }
+
+ ret = lookup_symbol(change, r->symbol, &vals);
+ if (ret != OK) {
+ release_vals(&vals);
+ return ret;
+ }
+ /*
+ * Relocations for the oldaddr fields of patches must have
+ * been resolved via run-pre matching.
+ */
+ if (!singular(&vals) || (r->symbol->candidate_vals != NULL &&
+ r->howto->type == KSPLICE_HOWTO_RELOC_PATCH)) {
+ release_vals(&vals);
+ ksdebug(change, "Failed to find %s for reloc\n",
+ r->symbol->label);
+ return FAILED_TO_FIND;
+ }
+ sym_addr = list_entry(vals.next, struct candidate_val, list)->val;
+ release_vals(&vals);
+
+ ret = write_reloc_value(change, r, r->blank_addr,
+ r->howto->pcrel ? sym_addr - r->blank_addr :
+ sym_addr);
+ if (ret != OK)
+ return ret;
+
+ ksdebug(change, "reloc: %lx to %s+%lx (S=%lx ", r->blank_addr,
+ r->symbol->label, r->target_addend, sym_addr);
+ switch (r->howto->size) {
+ case 1:
+ ksdebug(change, "aft=%02x)\n", *(uint8_t *)r->blank_addr);
+ break;
+ case 2:
+ ksdebug(change, "aft=%04x)\n", *(uint16_t *)r->blank_addr);
+ break;
+ case 4:
+ ksdebug(change, "aft=%08x)\n", *(uint32_t *)r->blank_addr);
+ break;
+#if BITS_PER_LONG >= 64
+ case 8:
+ ksdebug(change, "aft=%016llx)\n", *(uint64_t *)r->blank_addr);
+ break;
+#endif /* BITS_PER_LONG */
+ default:
+ ksdebug(change, "Aborted. Invalid relocation size.\n");
+ return UNEXPECTED;
+ }
+
+ /*
+ * Create labelvals so that we can verify our choices in the
+ * second round of run-pre matching that considers data sections.
+ */
+ ret = create_labelval(change, r->symbol, sym_addr, VAL);
+ if (ret != OK)
+ return ret;
+
+ return add_dependency_on_address(change, sym_addr);
+}
+
+/*
+ * Date relocations are created wherever __DATE__ or __TIME__ is used
+ * in the kernel; we resolve them by simply copying in the date/time
+ * obtained from run-pre matching the relevant compilation unit.
+ */
+static abort_t apply_howto_date(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r)
+{
+ if (r->symbol->candidate_vals != NULL) {
+ ksdebug(change, "Failed to find %s for date\n",
+ r->symbol->label);
+ return FAILED_TO_FIND;
+ }
+ memcpy((unsigned char *)r->blank_addr,
+ (const unsigned char *)r->symbol->value, r->howto->size);
+ return OK;
+}
+
+/*
+ * Given a relocation and its run address, compute the address of the
+ * symbol the relocation referenced, and store it in *valp.
+ */
+static abort_t read_reloc_value(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r,
+ unsigned long addr, unsigned long *valp)
+{
+ unsigned char bytes[sizeof(long)];
+ unsigned long val;
+ const struct ksplice_reloc_howto *howto = r->howto;
+
+ if (howto->size <= 0 || howto->size > sizeof(long)) {
+ ksdebug(change, "Aborted. Invalid relocation size.\n");
+ return UNEXPECTED;
+ }
+
+ if (probe_kernel_read(bytes, (void *)addr, howto->size) == -EFAULT)
+ return NO_MATCH;
+
+ switch (howto->size) {
+ case 1:
+ val = *(uint8_t *)bytes;
+ break;
+ case 2:
+ val = *(uint16_t *)bytes;
+ break;
+ case 4:
+ val = *(uint32_t *)bytes;
+ break;
+#if BITS_PER_LONG >= 64
+ case 8:
+ val = *(uint64_t *)bytes;
+ break;
+#endif /* BITS_PER_LONG */
+ default:
+ ksdebug(change, "Aborted. Invalid relocation size.\n");
+ return UNEXPECTED;
+ }
+
+ val &= howto->dst_mask;
+ if (howto->signed_addend)
+ val |= -(val & (howto->dst_mask & ~(howto->dst_mask >> 1)));
+ val <<= howto->rightshift;
+ val -= r->insn_addend + r->target_addend;
+ *valp = val;
+ return OK;
+}
+
+/*
+ * Given a relocation, the address of its storage unit, and the
+ * address of the symbol the relocation references, write the
+ * relocation's final value into the storage unit.
+ */
+static abort_t write_reloc_value(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r,
+ unsigned long addr, unsigned long sym_addr)
+{
+ unsigned long val = sym_addr + r->target_addend + r->insn_addend;
+ const struct ksplice_reloc_howto *howto = r->howto;
+ val >>= howto->rightshift;
+ switch (howto->size) {
+ case 1:
+ *(uint8_t *)addr = (*(uint8_t *)addr & ~howto->dst_mask) |
+ (val & howto->dst_mask);
+ break;
+ case 2:
+ *(uint16_t *)addr = (*(uint16_t *)addr & ~howto->dst_mask) |
+ (val & howto->dst_mask);
+ break;
+ case 4:
+ *(uint32_t *)addr = (*(uint32_t *)addr & ~howto->dst_mask) |
+ (val & howto->dst_mask);
+ break;
+#if BITS_PER_LONG >= 64
+ case 8:
+ *(uint64_t *)addr = (*(uint64_t *)addr & ~howto->dst_mask) |
+ (val & howto->dst_mask);
+ break;
+#endif /* BITS_PER_LONG */
+ default:
+ ksdebug(change, "Aborted. Invalid relocation size.\n");
+ return UNEXPECTED;
+ }
+
+ if (read_reloc_value(change, r, addr, &val) != OK || val != sym_addr) {
+ ksdebug(change, "Aborted. Relocation overflow.\n");
+ return UNEXPECTED;
+ }
+
+ return OK;
+}
+
+static abort_t create_module_list_entry(struct ksplice_mod_change *change,
+ bool to_be_applied)
+{
+ struct ksplice_module_list_entry *entry =
+ kmalloc(sizeof(*entry), GFP_KERNEL);
+ if (entry == NULL)
+ return OUT_OF_MEMORY;
+ entry->new_code_mod_name =
+ kstrdup(change->new_code_mod->name, GFP_KERNEL);
+ if (entry->new_code_mod_name == NULL) {
+ kfree(entry);
+ return OUT_OF_MEMORY;
+ }
+ entry->target_mod_name = kstrdup(change->target_name, GFP_KERNEL);
+ if (entry->target_mod_name == NULL) {
+ kfree(entry->new_code_mod_name);
+ kfree(entry);
+ return OUT_OF_MEMORY;
+ }
+ /* The update's kid is guaranteed to outlast the module_list_entry */
+ entry->kid = change->update->kid;
+ entry->applied = to_be_applied;
+ list_add(&entry->update_list, &change->update->ksplice_module_list);
+ return OK;
+}
+
+static void cleanup_module_list_entries(struct update *update)
+{
+ struct ksplice_module_list_entry *entry;
+ list_for_each_entry(entry, &update->ksplice_module_list, update_list) {
+ kfree(entry->target_mod_name);
+ kfree(entry->new_code_mod_name);
+ }
+ clear_list(&update->ksplice_module_list,
+ struct ksplice_module_list_entry, update_list);
+}
+
+/* Replacement address used for functions deleted by the patch */
+static void __attribute__((noreturn)) ksplice_deleted(void)
+{
+ printk(KERN_CRIT "Called a kernel function deleted by Ksplice!\n");
+ BUG();
+}
+
+/* Floodfill to run-pre match the sections within a change. */
+static abort_t match_change_sections(struct ksplice_mod_change *change,
+ bool consider_data_sections)
+{
+ struct ksplice_section *sect;
+ abort_t ret;
+ int remaining = 0;
+ bool progress;
+
+ for (sect = change->old_code.sections;
+ sect < change->old_code.sections_end; sect++) {
+ if ((sect->flags & KSPLICE_SECTION_DATA) == 0 &&
+ (sect->flags & KSPLICE_SECTION_STRING) == 0 &&
+ (sect->flags & KSPLICE_SECTION_MATCHED) == 0)
+ remaining++;
+ }
+
+ while (remaining > 0) {
+ progress = false;
+ for (sect = change->old_code.sections;
+ sect < change->old_code.sections_end; sect++) {
+ if ((sect->flags & KSPLICE_SECTION_MATCHED) != 0)
+ continue;
+ if ((!consider_data_sections &&
+ (sect->flags & KSPLICE_SECTION_DATA) != 0) ||
+ (sect->flags & KSPLICE_SECTION_STRING) != 0)
+ continue;
+ ret = find_section(change, sect);
+ if (ret == OK) {
+ sect->flags |= KSPLICE_SECTION_MATCHED;
+ if ((sect->flags & KSPLICE_SECTION_DATA) == 0)
+ remaining--;
+ progress = true;
+ } else if (ret != NO_MATCH) {
+ return ret;
+ }
+ }
+
+ if (progress)
+ continue;
+
+ for (sect = change->old_code.sections;
+ sect < change->old_code.sections_end; sect++) {
+ if ((sect->flags & KSPLICE_SECTION_MATCHED) != 0 ||
+ (sect->flags & KSPLICE_SECTION_STRING) != 0)
+ continue;
+ ksdebug(change, "run-pre: could not match %s "
+ "section %s\n",
+ (sect->flags & KSPLICE_SECTION_DATA) != 0 ?
+ "data" :
+ (sect->flags & KSPLICE_SECTION_RODATA) != 0 ?
+ "rodata" : "text", sect->symbol->label);
+ }
+ ksdebug(change, "Aborted. run-pre: could not match some "
+ "sections.\n");
+ return NO_MATCH;
+ }
+ return OK;
+}
+
+/*
+ * Search for the section in the running kernel. Returns OK if and
+ * only if it finds precisely one address in the kernel matching the
+ * section.
+ */
+static abort_t find_section(struct ksplice_mod_change *change,
+ struct ksplice_section *sect)
+{
+ int i;
+ abort_t ret;
+ unsigned long run_addr;
+ LIST_HEAD(vals);
+ struct candidate_val *v, *n;
+
+ ret = lookup_symbol(change, sect->symbol, &vals);
+ if (ret != OK) {
+ release_vals(&vals);
+ return ret;
+ }
+
+ ksdebug(change, "run-pre: starting sect search for %s\n",
+ sect->symbol->label);
+
+ list_for_each_entry_safe(v, n, &vals, list) {
+ run_addr = v->val;
+
+ yield();
+ ret = try_addr(change, sect, run_addr, NULL, RUN_PRE_INITIAL);
+ if (ret == NO_MATCH) {
+ list_del(&v->list);
+ kfree(v);
+ } else if (ret != OK) {
+ release_vals(&vals);
+ return ret;
+ }
+ }
+
+ if (singular(&vals)) {
+ LIST_HEAD(safety_records);
+ run_addr = list_entry(vals.next, struct candidate_val,
+ list)->val;
+ ret = try_addr(change, sect, run_addr, &safety_records,
+ RUN_PRE_FINAL);
+ release_vals(&vals);
+ if (ret != OK) {
+ clear_list(&safety_records, struct safety_record, list);
+ ksdebug(change, "run-pre: Final run failed for sect "
+ "%s:\n", sect->symbol->label);
+ } else {
+ list_splice(&safety_records, &change->safety_records);
+ }
+ return ret;
+ } else if (!list_empty(&vals)) {
+ struct candidate_val *val;
+ ksdebug(change, "run-pre: multiple candidates for sect %s:\n",
+ sect->symbol->label);
+ i = 0;
+ list_for_each_entry(val, &vals, list) {
+ i++;
+ ksdebug(change, "%lx\n", val->val);
+ if (i > 5) {
+ ksdebug(change, "...\n");
+ break;
+ }
+ }
+ release_vals(&vals);
+ return NO_MATCH;
+ }
+ release_vals(&vals);
+ return NO_MATCH;
+}
+
+/*
+ * try_addr is the the interface to run-pre matching. Its primary
+ * purpose is to manage debugging information for run-pre matching;
+ * all the hard work is in run_pre_cmp.
+ */
+static abort_t try_addr(struct ksplice_mod_change *change,
+ struct ksplice_section *sect,
+ unsigned long run_addr,
+ struct list_head *safety_records,
+ enum run_pre_mode mode)
+{
+ abort_t ret;
+ const struct module *run_module = __module_address(run_addr);
+
+ if (run_module == change->new_code_mod) {
+ ksdebug(change, "run-pre: unexpected address %lx in new_code "
+ "module %s for sect %s\n", run_addr, run_module->name,
+ sect->symbol->label);
+ return UNEXPECTED;
+ }
+ if (!patches_module(run_module, change->target)) {
+ ksdebug(change, "run-pre: ignoring address %lx in other module "
+ "%s for sect %s\n", run_addr, run_module == NULL ?
+ "vmlinux" : run_module->name, sect->symbol->label);
+ return NO_MATCH;
+ }
+
+ ret = create_labelval(change, sect->symbol, run_addr, TEMP);
+ if (ret != OK)
+ return ret;
+
+ ret = run_pre_cmp(change, sect, run_addr, safety_records, mode);
+ if (ret == NO_MATCH && mode != RUN_PRE_FINAL) {
+ set_temp_labelvals(change, NOVAL);
+ ksdebug(change, "run-pre: %s sect %s does not match (r_a=%lx "
+ "p_a=%lx s=%lx)\n",
+ (sect->flags & KSPLICE_SECTION_RODATA) != 0 ? "rodata" :
+ (sect->flags & KSPLICE_SECTION_DATA) != 0 ? "data" :
+ "text", sect->symbol->label, run_addr, sect->address,
+ sect->size);
+ ksdebug(change, "run-pre: ");
+ if (change->update->debug >= 1) {
+ ret = run_pre_cmp(change, sect, run_addr,
+ safety_records, RUN_PRE_DEBUG);
+ set_temp_labelvals(change, NOVAL);
+ }
+ ksdebug(change, "\n");
+ return ret;
+ } else if (ret != OK) {
+ set_temp_labelvals(change, NOVAL);
+ return ret;
+ }
+
+ if (mode != RUN_PRE_FINAL) {
+ set_temp_labelvals(change, NOVAL);
+ ksdebug(change, "run-pre: candidate for sect %s=%lx\n",
+ sect->symbol->label, run_addr);
+ return OK;
+ }
+
+ set_temp_labelvals(change, VAL);
+ ksdebug(change, "run-pre: found sect %s=%lx\n", sect->symbol->label,
+ run_addr);
+ return OK;
+}
+
+/*
+ * run_pre_cmp is the primary run-pre matching function; it determines
+ * whether the given ksplice_section matches the code or data in the
+ * running kernel starting at run_addr.
+ *
+ * If run_pre_mode is RUN_PRE_FINAL, a safety record for the matched
+ * section is created.
+ *
+ * The run_pre_mode is also used to determine what debugging
+ * information to display.
+ */
+static abort_t run_pre_cmp(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect,
+ unsigned long run_addr,
+ struct list_head *safety_records,
+ enum run_pre_mode mode)
+{
+ int matched = 0;
+ abort_t ret;
+ const struct ksplice_reloc *r, *finger;
+ const unsigned char *pre, *run, *pre_start, *run_start;
+ unsigned char runval;
+
+ pre_start = (const unsigned char *)sect->address;
+ run_start = (const unsigned char *)run_addr;
+
+ finger = init_reloc_search(change, sect);
+
+ pre = pre_start;
+ run = run_start;
+ while (pre < pre_start + sect->size) {
+ unsigned long offset = pre - pre_start;
+ ret = lookup_reloc(change, &finger, (unsigned long)pre, &r);
+ if (ret == OK) {
+ ret = handle_reloc(change, sect, r, (unsigned long)run,
+ mode);
+ if (ret != OK) {
+ if (mode == RUN_PRE_INITIAL)
+ ksdebug(change, "reloc in sect does "
+ "not match after %lx/%lx "
+ "bytes\n", offset, sect->size);
+ return ret;
+ }
+ if (mode == RUN_PRE_DEBUG)
+ print_bytes(change, run, r->howto->size, pre,
+ r->howto->size);
+ pre += r->howto->size;
+ run += r->howto->size;
+ finger++;
+ continue;
+ } else if (ret != NO_MATCH) {
+ return ret;
+ }
+
+ if ((sect->flags & KSPLICE_SECTION_TEXT) != 0) {
+ ret = handle_paravirt(change, (unsigned long)pre,
+ (unsigned long)run, &matched);
+ if (ret != OK)
+ return ret;
+ if (matched != 0) {
+ if (mode == RUN_PRE_DEBUG)
+ print_bytes(change, run, matched, pre,
+ matched);
+ pre += matched;
+ run += matched;
+ continue;
+ }
+ }
+
+ if (probe_kernel_read(&runval, (void *)run, 1) == -EFAULT) {
+ if (mode == RUN_PRE_INITIAL)
+ ksdebug(change, "sect unmapped after %lx/%lx "
+ "bytes\n", offset, sect->size);
+ return NO_MATCH;
+ }
+
+ if (runval != *pre &&
+ (sect->flags & KSPLICE_SECTION_DATA) == 0) {
+ if (mode == RUN_PRE_INITIAL)
+ ksdebug(change, "sect does not match after "
+ "%lx/%lx bytes\n", offset, sect->size);
+ if (mode == RUN_PRE_DEBUG) {
+ print_bytes(change, run, 1, pre, 1);
+ ksdebug(change, "[p_o=%lx] ! ", offset);
+ print_bytes(change, run + 1, 2, pre + 1, 2);
+ }
+ return NO_MATCH;
+ }
+ if (mode == RUN_PRE_DEBUG)
+ print_bytes(change, run, 1, pre, 1);
+ pre++;
+ run++;
+ }
+ return create_safety_record(change, sect, safety_records, run_addr,
+ run - run_start);
+}
+
+static void print_bytes(struct ksplice_mod_change *change,
+ const unsigned char *run, int runc,
+ const unsigned char *pre, int prec)
+{
+ int o;
+ int matched = min(runc, prec);
+ for (o = 0; o < matched; o++) {
+ if (run[o] == pre[o])
+ ksdebug(change, "%02x ", run[o]);
+ else
+ ksdebug(change, "%02x/%02x ", run[o], pre[o]);
+ }
+ for (o = matched; o < runc; o++)
+ ksdebug(change, "%02x/ ", run[o]);
+ for (o = matched; o < prec; o++)
+ ksdebug(change, "/%02x ", pre[o]);
+}
+
+struct range {
+ unsigned long address;
+ unsigned long size;
+};
+
+static int reloc_bsearch_compare(const void *key, const void *elt)
+{
+ const struct range *range = key;
+ const struct ksplice_reloc *r = elt;
+ if (range->address + range->size <= r->blank_addr)
+ return -1;
+ if (range->address > r->blank_addr)
+ return 1;
+ return 0;
+}
+
+static const struct ksplice_reloc *find_reloc(const struct ksplice_reloc *start,
+ const struct ksplice_reloc *end,
+ unsigned long address,
+ unsigned long size)
+{
+ const struct ksplice_reloc *r;
+ struct range range = { address, size };
+ r = bsearch((void *)&range, start, end - start, sizeof(*r),
+ reloc_bsearch_compare);
+ if (r == NULL)
+ return NULL;
+ while (r > start && (r - 1)->blank_addr >= address)
+ r--;
+ return r;
+}
+
+static const struct ksplice_reloc *
+init_reloc_search(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect)
+{
+ const struct ksplice_reloc *r;
+ r = find_reloc(change->old_code.relocs, change->old_code.relocs_end,
+ sect->address, sect->size);
+ if (r == NULL)
+ return change->old_code.relocs_end;
+ return r;
+}
+
+/*
+ * lookup_reloc implements an amortized O(1) lookup for the next
+ * old_code relocation. It must be called with a strictly increasing
+ * sequence of addresses.
+ *
+ * The fingerp is private data for lookup_reloc, and needs to have
+ * been initialized as a pointer to the result of find_reloc (or
+ * init_reloc_search).
+ */
+static abort_t lookup_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_reloc **fingerp,
+ unsigned long addr,
+ const struct ksplice_reloc **relocp)
+{
+ const struct ksplice_reloc *r = *fingerp;
+ int canary_ret;
+
+ while (r < change->old_code.relocs_end &&
+ addr >= r->blank_addr + r->howto->size &&
+ !(addr == r->blank_addr && r->howto->size == 0))
+ r++;
+ *fingerp = r;
+ if (r == change->old_code.relocs_end)
+ return NO_MATCH;
+ if (addr < r->blank_addr)
+ return NO_MATCH;
+ *relocp = r;
+ if (r->howto->type != KSPLICE_HOWTO_RELOC)
+ return OK;
+
+ canary_ret = contains_canary(change, r->blank_addr, r->howto);
+ if (canary_ret < 0)
+ return UNEXPECTED;
+ if (canary_ret == 0) {
+ ksdebug(change, "run-pre: reloc skipped at p_a=%lx to %s+%lx "
+ "(altinstr)\n", r->blank_addr, r->symbol->label,
+ r->target_addend);
+ return NO_MATCH;
+ }
+ if (addr != r->blank_addr) {
+ ksdebug(change, "Invalid nonzero relocation offset\n");
+ return UNEXPECTED;
+ }
+ return OK;
+}
+
+static abort_t handle_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect,
+ const struct ksplice_reloc *r,
+ unsigned long run_addr, enum run_pre_mode mode)
+{
+ switch (r->howto->type) {
+ case KSPLICE_HOWTO_RELOC:
+ return handle_howto_reloc(change, sect, r, run_addr, mode);
+ case KSPLICE_HOWTO_DATE:
+ case KSPLICE_HOWTO_TIME:
+ return handle_howto_date(change, sect, r, run_addr, mode);
+#ifdef CONFIG_BUG
+ case KSPLICE_HOWTO_BUG:
+ return handle_bug(change, r, run_addr);
+#endif /* CONFIG_BUG */
+ case KSPLICE_HOWTO_EXTABLE:
+ return handle_extable(change, r, run_addr);
+ default:
+ ksdebug(change, "Unexpected howto type %d\n", r->howto->type);
+ return UNEXPECTED;
+ }
+}
+
+/*
+ * For date/time relocations, we check that the sequence of bytes
+ * matches the format of a date or time.
+ */
+static abort_t handle_howto_date(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect,
+ const struct ksplice_reloc *r,
+ unsigned long run_addr, enum run_pre_mode mode)
+{
+ abort_t ret;
+ char *buf = kmalloc(r->howto->size, GFP_KERNEL);
+
+ if (buf == NULL)
+ return OUT_OF_MEMORY;
+ if (probe_kernel_read(buf, (void *)run_addr, r->howto->size) == -EFAULT) {
+ ret = NO_MATCH;
+ goto out;
+ }
+
+ switch (r->howto->type) {
+ case KSPLICE_HOWTO_TIME:
+ if (isdigit(buf[0]) && isdigit(buf[1]) && buf[2] == ':' &&
+ isdigit(buf[3]) && isdigit(buf[4]) && buf[5] == ':' &&
+ isdigit(buf[6]) && isdigit(buf[7]))
+ ret = OK;
+ else
+ ret = NO_MATCH;
+ break;
+ case KSPLICE_HOWTO_DATE:
+ if (isalpha(buf[0]) && isalpha(buf[1]) && isalpha(buf[2]) &&
+ buf[3] == ' ' && (buf[4] == ' ' || isdigit(buf[4])) &&
+ isdigit(buf[5]) && buf[6] == ' ' && isdigit(buf[7]) &&
+ isdigit(buf[8]) && isdigit(buf[9]) && isdigit(buf[10]))
+ ret = OK;
+ else
+ ret = NO_MATCH;
+ break;
+ default:
+ ret = UNEXPECTED;
+ }
+ if (ret == NO_MATCH && mode == RUN_PRE_INITIAL)
+ ksdebug(change, "%s string: \"%.*s\" does not match format\n",
+ r->howto->type == KSPLICE_HOWTO_DATE ? "date" : "time",
+ r->howto->size, buf);
+
+ if (ret != OK)
+ goto out;
+ ret = create_labelval(change, r->symbol, run_addr, TEMP);
+out:
+ kfree(buf);
+ return ret;
+}
+
+/*
+ * Extract the value of a symbol used in a relocation in the pre code
+ * during run-pre matching, giving an error if it conflicts with a
+ * previously found value of that symbol
+ */
+static abort_t handle_howto_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect,
+ const struct ksplice_reloc *r,
+ unsigned long run_addr,
+ enum run_pre_mode mode)
+{
+ struct ksplice_section *sym_sect = symbol_section(change, r->symbol);
+ unsigned long offset = r->target_addend;
+ unsigned long val;
+ abort_t ret;
+
+ ret = read_reloc_value(change, r, run_addr, &val);
+ if (ret != OK)
+ return ret;
+ if (r->howto->pcrel)
+ val += run_addr;
+
+ if (mode == RUN_PRE_INITIAL)
+ ksdebug(change, "run-pre: reloc at r_a=%lx p_a=%lx to %s+%lx: "
+ "found %s = %lx\n", run_addr, r->blank_addr,
+ r->symbol->label, offset, r->symbol->label, val);
+
+ if (contains_canary(change, run_addr, r->howto) != 0) {
+ ksdebug(change, "Aborted. Unexpected canary in run code at %lx"
+ "\n", run_addr);
+ return UNEXPECTED;
+ }
+
+ if ((sect->flags & KSPLICE_SECTION_DATA) != 0 &&
+ sect->symbol == r->symbol)
+ return OK;
+ ret = create_labelval(change, r->symbol, val, TEMP);
+ if (ret == NO_MATCH && mode == RUN_PRE_INITIAL)
+ ksdebug(change, "run-pre: reloc at r_a=%lx p_a=%lx: labelval "
+ "%s = %lx does not match expected %lx\n", run_addr,
+ r->blank_addr, r->symbol->label, r->symbol->value, val);
+
+ if (ret != OK)
+ return ret;
+ if (sym_sect != NULL && (sym_sect->flags & KSPLICE_SECTION_MATCHED) == 0
+ && (sym_sect->flags & KSPLICE_SECTION_STRING) != 0) {
+ if (mode == RUN_PRE_INITIAL)
+ ksdebug(change, "Recursively comparing string section "
+ "%s\n", sym_sect->symbol->label);
+ else if (mode == RUN_PRE_DEBUG)
+ ksdebug(change, "[str start] ");
+ ret = run_pre_cmp(change, sym_sect, val, NULL, mode);
+ if (mode == RUN_PRE_DEBUG)
+ ksdebug(change, "[str end] ");
+ if (ret == OK && mode == RUN_PRE_INITIAL)
+ ksdebug(change, "Successfully matched string section %s"
+ "\n", sym_sect->symbol->label);
+ else if (mode == RUN_PRE_INITIAL)
+ ksdebug(change, "Failed to match string section %s\n",
+ sym_sect->symbol->label);
+ }
+ return ret;
+}
+
+#ifdef CONFIG_GENERIC_BUG
+static abort_t handle_bug(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r, unsigned long run_addr)
+{
+ const struct bug_entry *run_bug = find_bug(run_addr);
+ struct ksplice_section *bug_sect = symbol_section(change, r->symbol);
+ if (run_bug == NULL)
+ return NO_MATCH;
+ if (bug_sect == NULL)
+ return UNEXPECTED;
+ return create_labelval(change, bug_sect->symbol, (unsigned long)run_bug,
+ TEMP);
+}
+#endif /* CONFIG_GENERIC_BUG */
+
+static abort_t handle_extable(struct ksplice_mod_change *change,
+ const struct ksplice_reloc *r,
+ unsigned long run_addr)
+{
+ const struct exception_table_entry *run_ent =
+ search_exception_tables(run_addr);
+ struct ksplice_section *ex_sect = symbol_section(change, r->symbol);
+ if (run_ent == NULL)
+ return NO_MATCH;
+ if (ex_sect == NULL)
+ return UNEXPECTED;
+ return create_labelval(change, ex_sect->symbol, (unsigned long)run_ent,
+ TEMP);
+}
+
+static int symbol_section_bsearch_compare(const void *a, const void *b)
+{
+ const struct ksplice_symbol *sym = a;
+ const struct ksplice_section *sect = b;
+ return strcmp(sym->label, sect->symbol->label);
+}
+
+static int compare_section_labels(const void *va, const void *vb)
+{
+ const struct ksplice_section *a = va, *b = vb;
+ return strcmp(a->symbol->label, b->symbol->label);
+}
+
+static struct ksplice_section *symbol_section(struct ksplice_mod_change *change,
+ const struct ksplice_symbol *sym)
+{
+ return bsearch(sym, change->old_code.sections,
+ change->old_code.sections_end -
+ change->old_code.sections,
+ sizeof(struct ksplice_section),
+ symbol_section_bsearch_compare);
+}
+
+/* Find the relocation for the oldaddr of a ksplice_patch */
+static const struct ksplice_reloc *
+patch_reloc(struct ksplice_mod_change *change,
+ const struct ksplice_patch *p)
+{
+ unsigned long addr = (unsigned long)&p->oldaddr;
+ const struct ksplice_reloc *r =
+ find_reloc(change->new_code.relocs, change->new_code.relocs_end,
+ addr, sizeof(addr));
+ if (r == NULL || r->blank_addr < addr ||
+ r->blank_addr >= addr + sizeof(addr))
+ return NULL;
+ return r;
+}
+
+/*
+ * Populates vals with the possible values for ksym from the various
+ * sources Ksplice uses to resolve symbols
+ */
+static abort_t lookup_symbol(struct ksplice_mod_change *change,
+ const struct ksplice_symbol *ksym,
+ struct list_head *vals)
+{
+ abort_t ret;
+
+ if (ksym->candidate_vals == NULL) {
+ release_vals(vals);
+ ksdebug(change, "using detected sym %s=%lx\n", ksym->label,
+ ksym->value);
+ return add_candidate_val(change, vals, ksym->value);
+ }
+
+ if (strcmp(ksym->label, "cleanup_module") == 0 && change->target != NULL
+ && change->target->exit != NULL) {
+ ret = add_candidate_val(change, vals,
+ (unsigned long)change->target->exit);
+ if (ret != OK)
+ return ret;
+ }
+
+ if (ksym->name != NULL) {
+ struct candidate_val *val;
+ list_for_each_entry(val, ksym->candidate_vals, list) {
+ ret = add_candidate_val(change, vals, val->val);
+ if (ret != OK)
+ return ret;
+ }
+
+ ret = new_export_lookup(change, ksym->name, vals);
+ if (ret != OK)
+ return ret;
+ }
+
+ return OK;
+}
+
+/*
+ * An update could one module to export a symbol and at the same time
+ * change another module to use that symbol. This violates the normal
+ * situation where the changes can be handled independently.
+ *
+ * new_export_lookup obtains symbol values from the changes to the
+ * exported symbol table made by other changes.
+ */
+static abort_t new_export_lookup(struct ksplice_mod_change *ichange,
+ const char *name, struct list_head *vals)
+{
+ struct ksplice_mod_change *change;
+ struct ksplice_patch *p;
+ list_for_each_entry(change, &ichange->update->changes, list) {
+ for (p = change->patches; p < change->patches_end; p++) {
+ const struct kernel_symbol *sym;
+ const struct ksplice_reloc *r;
+ if (p->type != KSPLICE_PATCH_EXPORT ||
+ strcmp(name, *(const char **)p->contents) != 0)
+ continue;
+
+ /* Check that the p->oldaddr reloc has been resolved. */
+ r = patch_reloc(change, p);
+ if (r == NULL ||
+ contains_canary(change, r->blank_addr,
+ r->howto) != 0)
+ continue;
+ sym = (const struct kernel_symbol *)r->symbol->value;
+
+ /*
+ * Check that the sym->value reloc has been resolved,
+ * if there is a Ksplice relocation there.
+ */
+ r = find_reloc(change->new_code.relocs,
+ change->new_code.relocs_end,
+ (unsigned long)&sym->value,
+ sizeof(&sym->value));
+ if (r != NULL &&
+ r->blank_addr == (unsigned long)&sym->value &&
+ contains_canary(change, r->blank_addr,
+ r->howto) != 0)
+ continue;
+ return add_candidate_val(ichange, vals, sym->value);
+ }
+ }
+ return OK;
+}
+
+/*
+ * When patch_action is called, the update should be fully prepared.
+ * patch_action will try to actually insert or remove trampolines for
+ * the update.
+ */
+static abort_t patch_action(struct update *update, enum ksplice_action action)
+{
+ static int (*const __patch_actions[KS_ACTIONS])(void *) = {
+ [KS_APPLY] = __apply_patches,
+ [KS_REVERSE] = __reverse_patches,
+ };
+ int i;
+ abort_t ret;
+ struct ksplice_mod_change *change;
+
+ ret = map_trampoline_pages(update);
+ if (ret != OK)
+ return ret;
+
+ list_for_each_entry(change, &update->changes, list) {
+ const typeof(int (*)(void)) *f;
+ for (f = change->hooks[action].pre;
+ f < change->hooks[action].pre_end; f++) {
+ if ((*f)() != 0) {
+ ret = CALL_FAILED;
+ goto out;
+ }
+ }
+ }
+
+ for (i = 0; i < 5; i++) {
+ cleanup_conflicts(update);
+ ret = (__force abort_t)stop_machine(__patch_actions[action],
+ update, NULL);
+ if (ret != CODE_BUSY)
+ break;
+ set_current_state(TASK_INTERRUPTIBLE);
+ schedule_timeout(msecs_to_jiffies(1000));
+ }
+out:
+ unmap_trampoline_pages(update);
+
+ if (ret == CODE_BUSY) {
+ print_conflicts(update);
+ _ksdebug(update, "Aborted %s. stack check: to-be-%s "
+ "code is busy.\n", update->kid,
+ action == KS_APPLY ? "replaced" : "reversed");
+ } else if (ret == ALREADY_REVERSED) {
+ _ksdebug(update, "Aborted %s. Ksplice update %s is already "
+ "reversed.\n", update->kid, update->kid);
+ } else if (ret == MODULE_BUSY) {
+ _ksdebug(update, "Update %s is in use by another module\n",
+ update->kid);
+ }
+
+ if (ret != OK) {
+ list_for_each_entry(change, &update->changes, list) {
+ const typeof(void (*)(void)) *f;
+ for (f = change->hooks[action].fail;
+ f < change->hooks[action].fail_end; f++)
+ (*f)();
+ }
+
+ return ret;
+ }
+
+ list_for_each_entry(change, &update->changes, list) {
+ const typeof(void (*)(void)) *f;
+ for (f = change->hooks[action].post;
+ f < change->hooks[action].post_end; f++)
+ (*f)();
+ }
+
+ _ksdebug(update, "Atomic patch %s for %s complete\n",
+ action == KS_APPLY ? "insertion" : "removal", update->kid);
+ return OK;
+}
+
+/* Atomically insert the update; run from within stop_machine */
+static int __apply_patches(void *updateptr)
+{
+ struct update *update = updateptr;
+ struct ksplice_mod_change *change;
+ struct ksplice_module_list_entry *entry;
+ struct ksplice_patch *p;
+ abort_t ret;
+
+ if (update->stage == STAGE_APPLIED)
+ return (__force int)OK;
+
+ if (update->stage != STAGE_PREPARING)
+ return (__force int)UNEXPECTED;
+
+ ret = check_each_task(update);
+ if (ret != OK)
+ return (__force int)ret;
+
+ list_for_each_entry(change, &update->changes, list) {
+ if (try_module_get(change->new_code_mod) != 1) {
+ struct ksplice_mod_change *change1;
+ list_for_each_entry(change1, &update->changes, list) {
+ if (change1 == change)
+ break;
+ module_put(change1->new_code_mod);
+ }
+ module_put(THIS_MODULE);
+ return (__force int)UNEXPECTED;
+ }
+ }
+
+ list_for_each_entry(change, &update->changes, list) {
+ const typeof(int (*)(void)) *f;
+ for (f = change->hooks[KS_APPLY].check;
+ f < change->hooks[KS_APPLY].check_end; f++) {
+ if ((*f)() != 0)
+ return (__force int)CALL_FAILED;
+ }
+ }
+
+ /* Commit point: the update application will succeed. */
+
+ update->stage = STAGE_APPLIED;
+ add_taint(TAINT_KSPLICE);
+
+ list_for_each_entry(entry, &update->ksplice_module_list, update_list)
+ list_add(&entry->list, &ksplice_modules);
+
+ list_for_each_entry(change, &update->changes, list) {
+ for (p = change->patches; p < change->patches_end; p++)
+ insert_trampoline(p);
+ }
+
+ list_for_each_entry(change, &update->changes, list) {
+ const typeof(void (*)(void)) *f;
+ for (f = change->hooks[KS_APPLY].intra;
+ f < change->hooks[KS_APPLY].intra_end; f++)
+ (*f)();
+ }
+
+ return (__force int)OK;
+}
+
+/* Atomically remove the update; run from within stop_machine */
+static int __reverse_patches(void *updateptr)
+{
+ struct update *update = updateptr;
+ struct ksplice_mod_change *change;
+ struct ksplice_module_list_entry *entry;
+ const struct ksplice_patch *p;
+ abort_t ret;
+
+ if (update->stage != STAGE_APPLIED)
+ return (__force int)OK;
+
+ list_for_each_entry(change, &update->changes, list) {
+ if (module_refcount(change->new_code_mod) != 1)
+ return (__force int)MODULE_BUSY;
+ }
+
+ list_for_each_entry(entry, &update->ksplice_module_list, update_list) {
+ if (!entry->applied &&
+ find_module(entry->target_mod_name) != NULL)
+ return COLD_UPDATE_LOADED;
+ }
+
+ ret = check_each_task(update);
+ if (ret != OK)
+ return (__force int)ret;
+
+ list_for_each_entry(change, &update->changes, list) {
+ for (p = change->patches; p < change->patches_end; p++) {
+ ret = verify_trampoline(change, p);
+ if (ret != OK)
+ return (__force int)ret;
+ }
+ }
+
+ list_for_each_entry(change, &update->changes, list) {
+ const typeof(int (*)(void)) *f;
+ for (f = change->hooks[KS_REVERSE].check;
+ f < change->hooks[KS_REVERSE].check_end; f++) {
+ if ((*f)() != 0)
+ return (__force int)CALL_FAILED;
+ }
+ }
+
+ /* Commit point: the update reversal will succeed. */
+
+ update->stage = STAGE_REVERSED;
+
+ list_for_each_entry(change, &update->changes, list)
+ module_put(change->new_code_mod);
+
+ list_for_each_entry(entry, &update->ksplice_module_list, update_list)
+ list_del(&entry->list);
+
+ list_for_each_entry(change, &update->changes, list) {
+ const typeof(void (*)(void)) *f;
+ for (f = change->hooks[KS_REVERSE].intra;
+ f < change->hooks[KS_REVERSE].intra_end; f++)
+ (*f)();
+ }
+
+ list_for_each_entry(change, &update->changes, list) {
+ for (p = change->patches; p < change->patches_end; p++)
+ remove_trampoline(p);
+ }
+
+ return (__force int)OK;
+}
+
+/*
+ * Check whether any thread's instruction pointer or any address of
+ * its stack is contained in one of the safety_records associated with
+ * the update.
+ *
+ * check_each_task must be called from inside stop_machine, because it
+ * does not take tasklist_lock (which cannot be held by anyone else
+ * during stop_machine).
+ */
+static abort_t check_each_task(struct update *update)
+{
+ const struct task_struct *g, *p;
+ abort_t status = OK, ret;
+ do_each_thread(g, p) {
+ /* do_each_thread is a double loop! */
+ ret = check_task(update, p, false);
+ if (ret != OK) {
+ check_task(update, p, true);
+ status = ret;
+ }
+ if (ret != OK && ret != CODE_BUSY)
+ return ret;
+ } while_each_thread(g, p);
+ return status;
+}
+
+static abort_t check_task(struct update *update,
+ const struct task_struct *t, bool rerun)
+{
+ abort_t status, ret;
+ struct conflict *conf = NULL;
+
+ if (rerun) {
+ conf = kmalloc(sizeof(*conf), GFP_ATOMIC);
+ if (conf == NULL)
+ return OUT_OF_MEMORY;
+ conf->process_name = kstrdup(t->comm, GFP_ATOMIC);
+ if (conf->process_name == NULL) {
+ kfree(conf);
+ return OUT_OF_MEMORY;
+ }
+ conf->pid = t->pid;
+ INIT_LIST_HEAD(&conf->stack);
+ list_add(&conf->list, &update->conflicts);
+ }
+
+ status = check_address(update, conf, KSPLICE_IP(t));
+ if (t == current) {
+ ret = check_stack(update, conf, task_thread_info(t),
+ (unsigned long *)__builtin_frame_address(0));
+ if (status == OK)
+ status = ret;
+ } else if (!task_curr(t)) {
+ ret = check_stack(update, conf, task_thread_info(t),
+ (unsigned long *)KSPLICE_SP(t));
+ if (status == OK)
+ status = ret;
+ } else if (!is_stop_machine(t)) {
+ status = UNEXPECTED_RUNNING_TASK;
+ }
+ return status;
+}
+
+static abort_t check_stack(struct update *update, struct conflict *conf,
+ const struct thread_info *tinfo,
+ const unsigned long *stack)
+{
+ abort_t status = OK, ret;
+ unsigned long addr;
+
+ while (valid_stack_ptr(tinfo, stack)) {
+ addr = *stack++;
+ ret = check_address(update, conf, addr);
+ if (ret != OK)
+ status = ret;
+ }
+ return status;
+}
+
+static abort_t check_address(struct update *update,
+ struct conflict *conf, unsigned long addr)
+{
+ abort_t status = OK, ret;
+ const struct safety_record *rec;
+ struct ksplice_mod_change *change;
+ struct conflict_addr *ca = NULL;
+
+ if (conf != NULL) {
+ ca = kmalloc(sizeof(*ca), GFP_ATOMIC);
+ if (ca == NULL)
+ return OUT_OF_MEMORY;
+ ca->addr = addr;
+ ca->has_conflict = false;
+ ca->label = NULL;
+ list_add(&ca->list, &conf->stack);
+ }
+
+ list_for_each_entry(change, &update->changes, list) {
+ unsigned long tramp_addr = follow_trampolines(change, addr);
+ list_for_each_entry(rec, &change->safety_records, list) {
+ ret = check_record(ca, rec, tramp_addr);
+ if (ret != OK)
+ status = ret;
+ }
+ }
+ return status;
+}
+
+static abort_t check_record(struct conflict_addr *ca,
+ const struct safety_record *rec, unsigned long addr)
+{
+ if (addr >= rec->addr && addr < rec->addr + rec->size) {
+ if (ca != NULL) {
+ ca->label = rec->label;
+ ca->has_conflict = true;
+ }
+ return CODE_BUSY;
+ }
+ return OK;
+}
+
+/* Is the task one of the stop_machine tasks? */
+static bool is_stop_machine(const struct task_struct *t)
+{
+ const char *kstop_prefix = "kstop/";
+ const char *num;
+ if (!strstarts(t->comm, kstop_prefix))
+ return false;
+ num = t->comm + strlen(kstop_prefix);
+ return num[strspn(num, "0123456789")] == '\0';
+}
+
+static void cleanup_conflicts(struct update *update)
+{
+ struct conflict *conf;
+ list_for_each_entry(conf, &update->conflicts, list) {
+ clear_list(&conf->stack, struct conflict_addr, list);
+ kfree(conf->process_name);
+ }
+ clear_list(&update->conflicts, struct conflict, list);
+}
+
+static void print_conflicts(struct update *update)
+{
+ const struct conflict *conf;
+ const struct conflict_addr *ca;
+ list_for_each_entry(conf, &update->conflicts, list) {
+ _ksdebug(update, "stack check: pid %d (%s):", conf->pid,
+ conf->process_name);
+ list_for_each_entry(ca, &conf->stack, list) {
+ _ksdebug(update, " %lx", ca->addr);
+ if (ca->has_conflict)
+ _ksdebug(update, " [<-CONFLICT]");
+ }
+ _ksdebug(update, "\n");
+ }
+}
+
+static void insert_trampoline(struct ksplice_patch *p)
+{
+ mm_segment_t old_fs = get_fs();
+ set_fs(KERNEL_DS);
+ memcpy(p->saved, p->vaddr, p->size);
+ memcpy(p->vaddr, p->contents, p->size);
+ flush_icache_range(p->oldaddr, p->oldaddr + p->size);
+ set_fs(old_fs);
+}
+
+static abort_t verify_trampoline(struct ksplice_mod_change *change,
+ const struct ksplice_patch *p)
+{
+ if (memcmp(p->vaddr, p->contents, p->size) != 0) {
+ ksdebug(change, "Aborted. Trampoline at %lx has been "
+ "overwritten.\n", p->oldaddr);
+ return CODE_BUSY;
+ }
+ return OK;
+}
+
+static void remove_trampoline(const struct ksplice_patch *p)
+{
+ mm_segment_t old_fs = get_fs();
+ set_fs(KERNEL_DS);
+ memcpy(p->vaddr, p->saved, p->size);
+ flush_icache_range(p->oldaddr, p->oldaddr + p->size);
+ set_fs(old_fs);
+}
+
+/* Returns NO_MATCH if there's already a labelval with a different value */
+static abort_t create_labelval(struct ksplice_mod_change *change,
+ struct ksplice_symbol *ksym,
+ unsigned long val, int status)
+{
+ val = follow_trampolines(change, val);
+ if (ksym->candidate_vals == NULL)
+ return ksym->value == val ? OK : NO_MATCH;
+
+ ksym->value = val;
+ if (status == TEMP) {
+ struct labelval *lv = kmalloc(sizeof(*lv), GFP_KERNEL);
+ if (lv == NULL)
+ return OUT_OF_MEMORY;
+ lv->symbol = ksym;
+ lv->saved_vals = ksym->candidate_vals;
+ list_add(&lv->list, &change->temp_labelvals);
+ }
+ ksym->candidate_vals = NULL;
+ return OK;
+}
+
+/*
+ * Creates a new safety_record for a old_code section based on its
+ * ksplice_section and run-pre matching information.
+ */
+static abort_t create_safety_record(struct ksplice_mod_change *change,
+ const struct ksplice_section *sect,
+ struct list_head *record_list,
+ unsigned long run_addr,
+ unsigned long run_size)
+{
+ struct safety_record *rec;
+ struct ksplice_patch *p;
+
+ if (record_list == NULL)
+ return OK;
+
+ for (p = change->patches; p < change->patches_end; p++) {
+ const struct ksplice_reloc *r = patch_reloc(change, p);
+ if (strcmp(sect->symbol->label, r->symbol->label) == 0)
+ break;
+ }
+ if (p >= change->patches_end)
+ return OK;
+
+ rec = kmalloc(sizeof(*rec), GFP_KERNEL);
+ if (rec == NULL)
+ return OUT_OF_MEMORY;
+ /*
+ * The old_code might be unloaded when checking reversing
+ * patches, so we need to kstrdup the label here.
+ */
+ rec->label = kstrdup(sect->symbol->label, GFP_KERNEL);
+ if (rec->label == NULL) {
+ kfree(rec);
+ return OUT_OF_MEMORY;
+ }
+ rec->addr = run_addr;
+ rec->size = run_size;
+
+ list_add(&rec->list, record_list);
+ return OK;
+}
+
+static abort_t add_candidate_val(struct ksplice_mod_change *change,
+ struct list_head *vals, unsigned long val)
+{
+ struct candidate_val *tmp, *new;
+
+/*
+ * Careful: follow trampolines before comparing values so that we do
+ * not mistake the obsolete function for another copy of the function.
+ */
+ val = follow_trampolines(change, val);
+
+ list_for_each_entry(tmp, vals, list) {
+ if (tmp->val == val)
+ return OK;
+ }
+ new = kmalloc(sizeof(*new), GFP_KERNEL);
+ if (new == NULL)
+ return OUT_OF_MEMORY;
+ new->val = val;
+ list_add(&new->list, vals);
+ return OK;
+}
+
+static void release_vals(struct list_head *vals)
+{
+ clear_list(vals, struct candidate_val, list);
+}
+
+/*
+ * The temp_labelvals list is used to cache those temporary labelvals
+ * that have been created to cross-check the symbol values obtained
+ * from different relocations within a single section being matched.
+ *
+ * If status is VAL, commit the temp_labelvals as final values.
+ *
+ * If status is NOVAL, restore the list of possible values to the
+ * ksplice_symbol, so that it no longer has a known value.
+ */
+static void set_temp_labelvals(struct ksplice_mod_change *change, int status)
+{
+ struct labelval *lv, *n;
+ list_for_each_entry_safe(lv, n, &change->temp_labelvals, list) {
+ if (status == NOVAL) {
+ lv->symbol->candidate_vals = lv->saved_vals;
+ } else {
+ release_vals(lv->saved_vals);
+ kfree(lv->saved_vals);
+ }
+ list_del(&lv->list);
+ kfree(lv);
+ }
+}
+
+/* Is there a Ksplice canary with given howto at blank_addr? */
+static int contains_canary(struct ksplice_mod_change *change,
+ unsigned long blank_addr,
+ const struct ksplice_reloc_howto *howto)
+{
+ switch (howto->size) {
+ case 1:
+ return (*(uint8_t *)blank_addr & howto->dst_mask) ==
+ (KSPLICE_CANARY & howto->dst_mask);
+ case 2:
+ return (*(uint16_t *)blank_addr & howto->dst_mask) ==
+ (KSPLICE_CANARY & howto->dst_mask);
+ case 4:
+ return (*(uint32_t *)blank_addr & howto->dst_mask) ==
+ (KSPLICE_CANARY & howto->dst_mask);
+#if BITS_PER_LONG >= 64
+ case 8:
+ return (*(uint64_t *)blank_addr & howto->dst_mask) ==
+ (KSPLICE_CANARY & howto->dst_mask);
+#endif /* BITS_PER_LONG */
+ default:
+ ksdebug(change, "Aborted. Invalid relocation size.\n");
+ return -1;
+ }
+}
+
+/*
+ * Compute the address of the code you would actually run if you were
+ * to call the function at addr (i.e., follow the sequence of jumps
+ * starting at addr)
+ */
+static unsigned long follow_trampolines(struct ksplice_mod_change *change,
+ unsigned long addr)
+{
+ unsigned long new_addr;
+ struct module *m;
+
+ while (1) {
+ if (!__kernel_text_address(addr) ||
+ trampoline_target(change, addr, &new_addr) != OK)
+ return addr;
+ m = __module_text_address(new_addr);
+ if (m == NULL || m == change->target ||
+ !strstarts(m->name, "ksplice"))
+ return addr;
+ addr = new_addr;
+ }
+}
+
+/* Does module a patch module b? */
+static bool patches_module(const struct module *a, const struct module *b)
+{
+ struct ksplice_module_list_entry *entry;
+ if (a == b)
+ return true;
+ list_for_each_entry(entry, &ksplice_modules, list) {
+ if (strcmp(entry->target_mod_name, b->name) == 0 &&
+ strcmp(entry->new_code_mod_name, a->name) == 0)
+ return true;
+ }
+ return false;
+}
+
+static bool singular(struct list_head *list)
+{
+ return !list_empty(list) && list->next->next == list;
+}
+
+static void *bsearch(const void *key, const void *base, size_t n,
+ size_t size, int (*cmp)(const void *key, const void *elt))
+{
+ int start = 0, end = n - 1, mid, result;
+ if (n == 0)
+ return NULL;
+ while (start <= end) {
+ mid = (start + end) / 2;
+ result = cmp(key, base + mid * size);
+ if (result < 0)
+ end = mid - 1;
+ else if (result > 0)
+ start = mid + 1;
+ else
+ return (void *)base + mid * size;
+ }
+ return NULL;
+}
+
+static int compare_relocs(const void *a, const void *b)
+{
+ const struct ksplice_reloc *ra = a, *rb = b;
+ if (ra->blank_addr > rb->blank_addr)
+ return 1;
+ else if (ra->blank_addr < rb->blank_addr)
+ return -1;
+ else
+ return ra->howto->size - rb->howto->size;
+}
+
+#ifdef CONFIG_DEBUG_FS
+
+static abort_t init_debug_buf(struct update *update)
+{
+ update->debug_blob.size = 0;
+ update->debug_blob.data = NULL;
+ update->debugfs_dentry =
+ debugfs_create_blob(update->name, S_IFREG | S_IRUSR, NULL,
+ &update->debug_blob);
+ if (update->debugfs_dentry == NULL)
+ return OUT_OF_MEMORY;
+ return OK;
+}
+
+static void clear_debug_buf(struct update *update)
+{
+ if (update->debugfs_dentry == NULL)
+ return;
+ debugfs_remove(update->debugfs_dentry);
+ update->debugfs_dentry = NULL;
+ update->debug_blob.size = 0;
+ vfree(update->debug_blob.data);
+ update->debug_blob.data = NULL;
+}
+
+static int _ksdebug(struct update *update, const char *fmt, ...)
+{
+ va_list args;
+ unsigned long size, old_size, new_size;
+
+ if (update->debug == 0)
+ return 0;
+
+ /* size includes the trailing '\0' */
+ va_start(args, fmt);
+ size = 1 + vsnprintf(update->debug_blob.data, 0, fmt, args);
+ va_end(args);
+ old_size = update->debug_blob.size == 0 ? 0 :
+ max(PAGE_SIZE, roundup_pow_of_two(update->debug_blob.size));
+ new_size = update->debug_blob.size + size == 0 ? 0 :
+ max(PAGE_SIZE, roundup_pow_of_two(update->debug_blob.size + size));
+ if (new_size > old_size) {
+ char *buf = vmalloc(new_size);
+ if (buf == NULL)
+ return -ENOMEM;
+ memcpy(buf, update->debug_blob.data, update->debug_blob.size);
+ vfree(update->debug_blob.data);
+ update->debug_blob.data = buf;
+ }
+ va_start(args, fmt);
+ update->debug_blob.size += vsnprintf(update->debug_blob.data +
+ update->debug_blob.size,
+ size, fmt, args);
+ va_end(args);
+ return 0;
+}
+#else /* CONFIG_DEBUG_FS */
+static abort_t init_debug_buf(struct update *update)
+{
+ return OK;
+}
+
+static void clear_debug_buf(struct update *update)
+{
+ return;
+}
+
+static int _ksdebug(struct update *update, const char *fmt, ...)
+{
+ va_list args;
+
+ if (update->debug == 0)
+ return 0;
+
+ if (!update->debug_continue_line)
+ printk(KERN_DEBUG "ksplice: ");
+
+ va_start(args, fmt);
+ vprintk(fmt, args);
+ va_end(args);
+
+ update->debug_continue_line =
+ fmt[0] == '\0' || fmt[strlen(fmt) - 1] != '\n';
+ return 0;
+}
+#endif /* CONFIG_DEBUG_FS */
+
+struct update_attribute {
+ struct attribute attr;
+ ssize_t (*show)(struct update *update, char *buf);
+ ssize_t (*store)(struct update *update, const char *buf, size_t len);
+};
+
+static ssize_t update_attr_show(struct kobject *kobj, struct attribute *attr,
+ char *buf)
+{
+ struct update_attribute *attribute =
+ container_of(attr, struct update_attribute, attr);
+ struct update *update = container_of(kobj, struct update, kobj);
+ if (attribute->show == NULL)
+ return -EIO;
+ return attribute->show(update, buf);
+}
+
+static ssize_t update_attr_store(struct kobject *kobj, struct attribute *attr,
+ const char *buf, size_t len)
+{
+ struct update_attribute *attribute =
+ container_of(attr, struct update_attribute, attr);
+ struct update *update = container_of(kobj, struct update, kobj);
+ if (attribute->store == NULL)
+ return -EIO;
+ return attribute->store(update, buf, len);
+}
+
+static struct sysfs_ops update_sysfs_ops = {
+ .show = update_attr_show,
+ .store = update_attr_store,
+};
+
+static void update_release(struct kobject *kobj)
+{
+ struct update *update;
+ update = container_of(kobj, struct update, kobj);
+ cleanup_ksplice_update(update);
+}
+
+static ssize_t stage_show(struct update *update, char *buf)
+{
+ switch (update->stage) {
+ case STAGE_PREPARING:
+ return snprintf(buf, PAGE_SIZE, "preparing\n");
+ case STAGE_APPLIED:
+ return snprintf(buf, PAGE_SIZE, "applied\n");
+ case STAGE_REVERSED:
+ return snprintf(buf, PAGE_SIZE, "reversed\n");
+ }
+ return 0;
+}
+
+static ssize_t abort_cause_show(struct update *update, char *buf)
+{
+ switch (update->abort_cause) {
+ case OK:
+ return snprintf(buf, PAGE_SIZE, "ok\n");
+ case NO_MATCH:
+ return snprintf(buf, PAGE_SIZE, "no_match\n");
+ case CODE_BUSY:
+ return snprintf(buf, PAGE_SIZE, "code_busy\n");
+ case MODULE_BUSY:
+ return snprintf(buf, PAGE_SIZE, "module_busy\n");
+ case OUT_OF_MEMORY:
+ return snprintf(buf, PAGE_SIZE, "out_of_memory\n");
+ case FAILED_TO_FIND:
+ return snprintf(buf, PAGE_SIZE, "failed_to_find\n");
+ case ALREADY_REVERSED:
+ return snprintf(buf, PAGE_SIZE, "already_reversed\n");
+ case MISSING_EXPORT:
+ return snprintf(buf, PAGE_SIZE, "missing_export\n");
+ case UNEXPECTED_RUNNING_TASK:
+ return snprintf(buf, PAGE_SIZE, "unexpected_running_task\n");
+ case TARGET_NOT_LOADED:
+ return snprintf(buf, PAGE_SIZE, "target_not_loaded\n");
+ case CALL_FAILED:
+ return snprintf(buf, PAGE_SIZE, "call_failed\n");
+ case COLD_UPDATE_LOADED:
+ return snprintf(buf, PAGE_SIZE, "cold_update_loaded\n");
+ case UNEXPECTED:
+ return snprintf(buf, PAGE_SIZE, "unexpected\n");
+ default:
+ return snprintf(buf, PAGE_SIZE, "unknown\n");
+ }
+ return 0;
+}
+
+static ssize_t conflict_show(struct update *update, char *buf)
+{
+ const struct conflict *conf;
+ const struct conflict_addr *ca;
+ int used = 0;
+ mutex_lock(&module_mutex);
+ list_for_each_entry(conf, &update->conflicts, list) {
+ used += snprintf(buf + used, PAGE_SIZE - used, "%s %d",
+ conf->process_name, conf->pid);
+ list_for_each_entry(ca, &conf->stack, list) {
+ if (!ca->has_conflict)
+ continue;
+ used += snprintf(buf + used, PAGE_SIZE - used, " %s",
+ ca->label);
+ }
+ used += snprintf(buf + used, PAGE_SIZE - used, "\n");
+ }
+ mutex_unlock(&module_mutex);
+ return used;
+}
+
+/* Used to pass maybe_cleanup_ksplice_update to kthread_run */
+static int maybe_cleanup_ksplice_update_wrapper(void *updateptr)
+{
+ struct update *update = updateptr;
+ mutex_lock(&module_mutex);
+ maybe_cleanup_ksplice_update(update);
+ mutex_unlock(&module_mutex);
+ return 0;
+}
+
+static ssize_t stage_store(struct update *update, const char *buf, size_t len)
+{
+ enum stage old_stage;
+ mutex_lock(&module_mutex);
+ old_stage = update->stage;
+ if ((strncmp(buf, "applied", len) == 0 ||
+ strncmp(buf, "applied\n", len) == 0) &&
+ update->stage == STAGE_PREPARING)
+ update->abort_cause = apply_update(update);
+ else if ((strncmp(buf, "reversed", len) == 0 ||
+ strncmp(buf, "reversed\n", len) == 0) &&
+ update->stage == STAGE_APPLIED)
+ update->abort_cause = reverse_update(update);
+ else if ((strncmp(buf, "cleanup", len) == 0 ||
+ strncmp(buf, "cleanup\n", len) == 0) &&
+ update->stage == STAGE_REVERSED)
+ kthread_run(maybe_cleanup_ksplice_update_wrapper, update,
+ "ksplice_cleanup_%s", update->kid);
+
+ mutex_unlock(&module_mutex);
+ return len;
+}
+
+static ssize_t debug_show(struct update *update, char *buf)
+{
+ return snprintf(buf, PAGE_SIZE, "%d\n", update->debug);
+}
+
+static ssize_t debug_store(struct update *update, const char *buf, size_t len)
+{
+ unsigned long l;
+ int ret = strict_strtoul(buf, 10, &l);
+ if (ret != 0)
+ return ret;
+ update->debug = l;
+ return len;
+}
+
+static ssize_t partial_show(struct update *update, char *buf)
+{
+ return snprintf(buf, PAGE_SIZE, "%d\n", update->partial);
+}
+
+static ssize_t partial_store(struct update *update, const char *buf, size_t len)
+{
+ unsigned long l;
+ int ret = strict_strtoul(buf, 10, &l);
+ if (ret != 0)
+ return ret;
+ update->partial = l;
+ return len;
+}
+
+static struct update_attribute stage_attribute =
+ __ATTR(stage, 0600, stage_show, stage_store);
+static struct update_attribute abort_cause_attribute =
+ __ATTR(abort_cause, 0400, abort_cause_show, NULL);
+static struct update_attribute debug_attribute =
+ __ATTR(debug, 0600, debug_show, debug_store);
+static struct update_attribute partial_attribute =
+ __ATTR(partial, 0600, partial_show, partial_store);
+static struct update_attribute conflict_attribute =
+ __ATTR(conflicts, 0400, conflict_show, NULL);
+
+static struct attribute *update_attrs[] = {
+ &stage_attribute.attr,
+ &abort_cause_attribute.attr,
+ &debug_attribute.attr,
+ &partial_attribute.attr,
+ &conflict_attribute.attr,
+ NULL
+};
+
+static struct kobj_type update_ktype = {
+ .sysfs_ops = &update_sysfs_ops,
+ .release = update_release,
+ .default_attrs = update_attrs,
+};
+
+static int init_ksplice(void)
+{
+ ksplice_kobj = kobject_create_and_add("ksplice", kernel_kobj);
+ if (ksplice_kobj == NULL)
+ return -ENOMEM;
+ return 0;
+}
+
+static void cleanup_ksplice(void)
+{
+ kobject_put(ksplice_kobj);
+}
+
+module_init(init_ksplice);
+module_exit(cleanup_ksplice);
+
+MODULE_AUTHOR("Ksplice, Inc.");
+MODULE_DESCRIPTION("Ksplice rebootless update system");
+MODULE_LICENSE("GPL v2");
diff --git a/kernel/panic.c b/kernel/panic.c
index 2a2ff36..90faf4a 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -156,6 +156,7 @@ static const struct tnt tnts[] = {
{ TAINT_OVERRIDDEN_ACPI_TABLE, 'A', ' ' },
{ TAINT_WARN, 'W', ' ' },
{ TAINT_CRAP, 'C', ' ' },
+ { TAINT_KSPLICE, 'K', ' ' },
};
/**
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists