- 19 May, 2018 8 commits
-
-
Kirill A. Shutemov authored
__pgtable_l5_enabled shouldn't be needed after system has booted. All preparation is done. We can now mark it as __initdata. Signed-off-by:
Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Reviewed-by:
Thomas Gleixner <tglx@linutronix.de> Cc: Hugh Dickins <hughd@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/20180518103528.59260-8-kirill.shutemov@linux.intel.comSigned-off-by:
Ingo Molnar <mingo@kernel.org>
-
Kirill A. Shutemov authored
__pgtable_l5_enabled shouldn't be needed after system has booted, we can mark it as __initdata, but it requires preparation. KASAN initialization code is a user of USE_EARLY_PGTABLE_L5, so all pgtable_l5_enabled() translated to __pgtable_l5_enabled there, including the one in p4d_offset(). It may lead to section mismatch, if a compiler would not inline p4d_offset(), but leave it as a standalone function: p4d_offset() is not marked as __init. Marking p4d_offset() as __always_inline fixes the issue. Signed-off-by:
-
Kirill A. Shutemov authored
This kernel parameter allows to force kernel to use 4-level paging even if hardware and kernel support 5-level paging. The option may be useful to work around regressions related to 5-level paging. Signed-off-by:
-
Kirill A. Shutemov authored
pgtable_l5_enabled is defined using cpu_feature_enabled() but we refer to it as a variable. This is misleading. Make pgtable_l5_enabled() a function. We cannot literally define it as a function due to circular dependencies between header files. Function-alike macros is close enough. Signed-off-by:
-
Kirill A. Shutemov authored
Usually pgtable_l5_enabled is defined using cpu_feature_enabled(). cpu_feature_enabled() is not available in early boot code. We use several different preprocessor tricks to get around it. It's messy. Unify them all. If cpu_feature_enabled() is not yet available, USE_EARLY_PGTABLE_L5 can be defined before all includes. It makes pgtable_l5_enabled rely on __pgtable_l5_enabled variable instead. This approach fits all early users. Signed-off-by:
-
Kirill A. Shutemov authored
Hugh noticied that we calculate the address of the trampoline page table incorrectly in cleanup_trampoline(). TRAMPOLINE_32BIT_PGTABLE_OFFSET has to be divided by sizeof(unsigned long), since trampoline_32bit is an 'unsigned long' pointer. TRAMPOLINE_32BIT_PGTABLE_OFFSET is zero so the bug doesn't have a visible effect. Reported-by:
Hugh Dickins <hughd@google.com> Signed-off-by:
-
Ingo Molnar authored
Signed-off-by: -
Josh Poimboeuf authored
With the following commit: fd35c88b ("objtool: Support GCC 8 switch tables") I added a "can't find switch jump table" warning, to stop covering up silent failures if add_switch_table() can't find anything. That warning found yet another bug in the objtool switch table detection logic. For cases 1 and 2 (as described in the comments of find_switch_table()), the find_symbol_containing() check doesn't adjust the offset for RIP-relative switch jumps. Incidentally, this bug was already fixed for case 3 with: 6f5ec299 ("objtool: Detect RIP-relative switch table references") However, that commit missed the fix for cases 1 and 2. The different cases are now starting to look more and more alike. So fix the bug by consolidating them into a single case, by checking the original dynamic jump instruction in the case 3 loop. This also simplifies the code and makes it more robust against future switch table detection issues -- of which I'm sure there will be many... Switch table detection has been the most fragile area of objtool, by far. I long for the day when we'll have a GCC plugin for annotating switch tables. Linus asked me to delay such a plugin due to the flakiness of the plugin infrastructure in older versions of GCC, so this rickety code is what we're stuck with for now. At least the code is now a little simpler than it was. Reported-by:
kbuild test robot <lkp@intel.com> Signed-off-by:
Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/f400541613d45689086329432f3095119ffbc328.1526674218.git.jpoimboe@redhat.comSigned-off-by:
-
- 17 May, 2018 1 commit
-
-
Thomas Gleixner authored
Rick bisected a regression on large systems which use the x2apic cluster mode for interrupt delivery to the commit wich reworked the cluster management. The problem is caused by a missing initialization of the clusterid field in the shared cluster data structures. So all structures end up with cluster ID 0 which only allows sharing between all CPUs which belong to cluster 0. All other CPUs with a cluster ID > 0 cannot share the data structure because they cannot find existing data with their cluster ID. This causes malfunction with IPIs because IPIs are sent to the wrong cluster and the caller waits for ever that the target CPU handles the IPI. Add the missing initialization when a upcoming CPU is the first in a cluster so that the later booting CPUs can find the data and share it for proper operation. Fixes: 023a6117 ("x86/apic/x2apic: Simplify cluster management") Reported-by:
Rick Warner <rick@microway.com> Bisected-by:
-
- 16 May, 2018 2 commits
-
-
Kirill A. Shutemov authored
cleanup_trampoline() relocates the top-level page table out of trampoline memory. We use 'top_pgtable' as our new top-level page table. But if the 'top_pgtable' would be referenced from C in a usual way, the address of the table will be calculated relative to RIP. After kernel gets relocated, the address will be in the middle of decompression buffer and the page table may get overwritten. This leads to a crash. We calculate the address of other page tables relative to the relocation address. It makes them safe. We should do the same for 'top_pgtable'. Calculate the address of 'top_pgtable' in assembly and pass down to cleanup_trampoline(). Move the page table to .pgtable section where the rest of page tables are. The section is @nobits so we save 4k in kernel image. Signed-off-by:
-
Kirill A. Shutemov authored
Eric and Hugh have reported instant reboot due to my recent changes in decompression code. The root cause is that I didn't realize that we need to adjust GOT to be able to run C code that early. The problem is only visible with an older toolchain. Binutils >= 2.24 is able to eliminate GOT references by replacing them with RIP-relative address loads: https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commitdiff;h=80d873266dec We need to adjust GOT two times: - before calling paging_prepare() using the initial load address - before calling C code from the relocated kernel Reported-by:
Eric Dumazet <eric.dumazet@gmail.com> Reported-by:
-
- 15 May, 2018 1 commit
-
-
Josh Poimboeuf authored
Typically a switch table can be found by detecting a .rodata access followed an indirect jump: 1969: 4a 8b 0c e5 00 00 00 mov 0x0(,%r12,8),%rcx 1970: 00 196d: R_X86_64_32S .rodata+0x438 1971: e9 00 00 00 00 jmpq 1976 <dispc_runtime_suspend+0xb6a> 1972: R_X86_64_PC32 __x86_indirect_thunk_rcx-0x4 Randy Dunlap reported a case (seen with GCC 4.8) where the .rodata access uses RIP-relative addressing: 19bd: 48 8b 3d 00 00 00 00 mov 0x0(%rip),%rdi # 19c4 <dispc_runtime_suspend+0xbb8> 19c0: R_X86_64_PC32 .rodata+0x45c 19c4: e9 00 00 00 00 jmpq 19c9 <dispc_runtime_suspend+0xbbd> 19c5: R_X86_64_PC32 __x86_indirect_thunk_rdi-0x4 In this case the relocation addend needs to be adjusted accordingly in order to find the location of the switch table. The fix is for case 3 (as described in the comments), but also make the existing case 1 & 2 checks more precise by only adjusting the addend for R_X86_64_PC32 relocations. This fixes the following warnings: drivers/video/fbdev/omap2/omapfb/dss/dispc.o: warning: objtool: dispc_runtime_suspend()+0xbb8: sibling call from callable instruction with modified stack frame drivers/video/fbdev/omap2/omapfb/dss/dispc.o: warning: objtool: dispc_runtime_resume()+0xcc5: sibling call from callable instruction with modified stack frame Reported-by:Randy Dunlap <rdunlap@infradead.org> Signed-off-by:
-
- 14 May, 2018 21 commits
-
-
Dave Hansen authored
mm_pkey_is_allocated() treats pkey 0 as unallocated. That is inconsistent with the manpages, and also inconsistent with mm->context.pkey_allocation_map. Stop special casing it and only disallow values that are actually bad (< 0). The end-user visible effect of this is that you can now use mprotect_pkey() to set pkey=0. This is a bit nicer than what Ram proposed[1] because it is simpler and removes special-casing for pkey 0. On the other hand, it does allow applications to pkey_free() pkey-0, but that's just a silly thing to do, so we are not going to protect against it. The scenario that could happen is similar to what happens if you free any other pkey that is in use: it might get reallocated later and used to protect some other data. The most likely scenario is that pkey-0 comes back from pkey_alloc(), an access-disable or write-disable bit is set in PKRU for it, and the next stack access will SIGSEGV. It's not horribly different from if you mprotect()'d your stack or heap to be unreadable or unwritable, which is generally very foolish, but also not explicitly prevented by the kernel. 1. http://lkml.kernel.org/r/1522112702-27853-1-git-send-email-linuxram@us.ibm.comSigned-off-by:
Dave Hansen <dave.hansen@linux.intel.com> Cc: Andrew Morton <akpm@linux-foundation.org>p Cc: Dave Hansen <dave.hansen@intel.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michael Ellermen <mpe@ellerman.id.au> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Shuah Khan <shuah@kernel.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mm@kvack.org Cc: stable@vger.kernel.org Fixes: 58ab9a08 ("x86/pkeys: Check against max pkey to avoid overflows") Link: http://lkml.kernel.org/r/20180509171358.47FD785E@viggo.jf.intel.comSigned-off-by:
-
Dave Hansen authored
Protection key 0 is the default key for all memory and will not normally come back from pkey_alloc(). But, you might still want pass it to mprotect_pkey(). This check ensures that you can use pkey 0. Signed-off-by:
-
Dave Hansen authored
This makes it possible to to tell what 'prot' a given allocation is supposed to have. That way, if we want to change just the pkey, we know what 'prot' to pass to mprotect_pkey(). Also, keep a record of the most recent allocation so the tests can easily find it. Signed-off-by:
-
Dave Hansen authored
We dump out the entire area of the siginfo where the si_pkey_ptr is supposed to be. But, we do some math on the poitner, which is a u32. We intended to do byte math, not u32 math on the pointer. Cast it over to a u8* so it works. Also, move this block of code to below th si_code check. It doesn't hurt anything, but the si_pkey field is gibberish for other signal types. Signed-off-by:
-
Dave Hansen authored
I got a bug report that the following code (roughly) was causing a SIGSEGV: mprotect(ptr, size, PROT_EXEC); mprotect(ptr, size, PROT_NONE); mprotect(ptr, size, PROT_READ); *ptr = 100; The problem is hit when the mprotect(PROT_EXEC) is implicitly assigned a protection key to the VMA, and made that key ACCESS_DENY|WRITE_DENY. The PROT_NONE mprotect() failed to remove the protection key, and the PROT_NONE-> PROT_READ left the PTE usable, but the pkey still in place and left the memory inaccessible. To fix this, we ensure that we always "override" the pkee at mprotect() if the VMA does not have execute-only permissions, but the VMA has the execute-only pkey. We had a check for PROT_READ/WRITE, but it did not work for PROT_NONE. This entirely removes the PROT_* checks, which ensures that PROT_NONE now works. Reported-by:
Shakeel Butt <shakeelb@google.com> Signed-off-by:
-
Dave Hansen authored
In our "exhaust all pkeys" test, we make sure that there is the expected number available. Turns out that the test did not cover the execute-only key, but discussed it anyway. It did *not* discuss the test-allocated key. Now that we have a test for the mprotect(PROT_EXEC) case, this off-by-one issue showed itself. Correct the off-by- one and add the explanation for the case we missed. Signed-off-by:
-
Dave Hansen authored
Under the covers, implement executable-only memory with protection keys when userspace calls mprotect(PROT_EXEC). But, we did not have a selftest for that. Now we do. Signed-off-by:
-
Dave Hansen authored
We currently have an execute-only test, but it is for the explicit mprotect_pkey() interface. We will soon add a test for the implicit mprotect(PROT_EXEC) enterface. We need this code in both tests. Signed-off-by:
-
Dave Hansen authored
The exec-only pkey is allocated inside the kernel and userspace is not told what it is. So, allow PK faults to occur that have an unknown key. Signed-off-by:
-
Dave Hansen authored
printf() and friends are unusable in signal handlers. They deadlock. The pkey selftest does not do any normal printing in signal handlers, only extra debugging. So, just print the format string so we get *some* output when debugging. Signed-off-by:
-
Dave Hansen authored
There is some noisy debug code at the end of the signal handler. It was disabled by an early, unconditional "return". However, that return also hid a dprint_in_signal=0, which kept dprint_in_signal=1 and effectively locked us into permanent dprint_in_signal=1 behavior. Remove the return and the dead code, fixing dprint_in_signal. Signed-off-by:
-
Dave Hansen authored
If we use assert(), the program "crashes". That can be scary to users, so stop doing it. Just exit with a >0 exit code instead. Signed-off-by:
-
Dave Hansen authored
do_not_expect_pk_fault() is a helper that we call when we do not expect a PK fault to have occurred. But, it is a function, which means that it obscures the line numbers from pkey_assert(). It also gives no details. Replace it with an implementation that gives nice line numbers and also lets callers pass in a more descriptive message about what happened that caused the unexpected fault. Signed-off-by:
-
Andy Lutomirski authored
This exercises a nasty corner case of the x86 ISA. Signed-off-by:
Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/67e08b69817171da8026e0eb3af0214b06b4d74f.1525800455.git.luto@kernel.orgSigned-off-by:
-
Ingo Molnar authored
Fix this warning: mpx-mini-test.c:422:0: warning: "SEGV_BNDERR" redefined Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: akpm@linux-foundation.org Cc: dave.hansen@intel.com Cc: linux-mm@kvack.org Cc: linuxram@us.ibm.com Cc: mpe@ellerman.id.au Cc: shakeelb@google.com Cc: shuah@kernel.org Link: http://lkml.kernel.org/r/20180514085908.GA12798@gmail.comSigned-off-by:
-
Ingo Molnar authored
Ubuntu 18.04 started exporting pkeys details in header files, resulting in build failures and warnings in the pkeys self-tests: protection_keys.c:232:0: warning: "SEGV_BNDERR" redefined protection_keys.c:387:5: error: conflicting types for ‘pkey_get’ protection_keys.c:409:5: error: conflicting types for ‘pkey_set’ ... Fix these namespace conflicts and double definitions, plus also clean up the ABI definitions to make it all a bit more readable ... Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: akpm@linux-foundation.org Cc: dave.hansen@intel.com Cc: linux-mm@kvack.org Cc: linuxram@us.ibm.com Cc: mpe@ellerman.id.au Cc: shakeelb@google.com Cc: shuah@kernel.org Link: http://lkml.kernel.org/r/20180514085623.GB7094@gmail.comSigned-off-by:
-
Alexander Potapenko authored
Clang builds with defconfig started crashing after the following commit: fb43d6cb ("x86/mm: Do not auto-massage page protections") This was caused by introducing a new global access in __startup_64(). Code in __startup_64() can be relocated during execution, but the compiler doesn't have to generate PC-relative relocations when accessing globals from that function. Clang actually does not generate them, which leads to boot-time crashes. To work around this problem, every global pointer must be adjusted using fixup_pointer(). Signed-off-by:
Alexander Potapenko <glider@google.com> Reviewed-by:
Dave Hansen <dave.hansen@intel.com> Acked-by:
-
Josh Poimboeuf authored
With GCC 8, some issues were found with the objtool switch table detection. 1) In the .rodata section, immediately after the switch table, there can be another object which contains a pointer to the function which had the switch statement. In this case objtool wrongly considers the function pointer to be part of the switch table. Fix it by: a) making sure there are no pointers to the beginning of the function; and b) making sure there are no gaps in the switch table. Only the former was needed, the latter adds additional protection for future optimizations. 2) In find_switch_table(), case 1 and case 2 are missing the check to ensure that the .rodata switch table data is anonymous, i.e. that it isn't already associated with an ELF symbol. Fix it by adding the same find_symbol_containing() check which is used for case 3. This fixes the following warnings with GCC 8: drivers/block/virtio_blk.o: warning: objtool: virtio_queue_rq()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+72 net/ipv6/icmp.o: warning: objtool: icmpv6_rcv()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+64 drivers/usb/core/quirks.o: warning: objtool: quirks_param_set()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+48 drivers/mtd/nand/raw/nand_hynix.o: warning: objtool: hynix_nand_decode_id()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+24 drivers/mtd/nand/raw/nand_samsung.o: warning: objtool: samsung_nand_decode_id()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+32 drivers/gpu/drm/nouveau/nvkm/subdev/top/gk104.o: warning: objtool: gk104_top_oneinit()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+64 Reported-by:Arnd Bergmann <arnd@arndb.de> Reported-by:
Peter Zijlstra (Intel) <peterz@infradead.org> Cc: David Laight <David.Laight@ACULAB.COM> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: damian <damian.tometzki@icloud.com> Link: http://lkml.kernel.org/r/20180510224849.xwi34d6tzheb5wgw@trebleSigned-off-by:
-
Josh Poimboeuf authored
GCC 8 moves a lot of unlikely code out of line to "cold" subfunctions in .text.unlikely. Properly detect the new subfunctions and treat them as extensions of the original functions. This fixes a bunch of warnings like: kernel/cgroup/cgroup.o: warning: objtool: parse_cgroup_root_flags()+0x33: sibling call from callable instruction with modified stack frame kernel/cgroup/cgroup.o: warning: objtool: cgroup_addrm_files()+0x290: sibling call from callable instruction with modified stack frame kernel/cgroup/cgroup.o: warning: objtool: cgroup_apply_control_enable()+0x25b: sibling call from callable instruction with modified stack frame kernel/cgroup/cgroup.o: warning: objtool: rebind_subsystems()+0x325: sibling call from callable instruction with modified stack frame Reported-and-tested-by:
damian <damian.tometzki@icloud.com> Reported-by:
-
Josh Poimboeuf authored
Objtool has some crude logic for detecting static "noreturn" functions (aka "dead ends"). This is necessary for being able to correctly follow GCC code flow when such functions are called. It's remotely possible for two functions to call each other via sibling calls. If they don't have RET instructions, objtool's noreturn detection logic goes into a recursive loop: drivers/char/ipmi/ipmi_ssif.o: warning: objtool: return_hosed_msg()+0x0: infinite recursion (objtool bug!) drivers/char/ipmi/ipmi_ssif.o: warning: objtool: deliver_recv_msg()+0x0: infinite recursion (objtool bug!) Instead of reporting an error in this case, consider the functions to be non-dead-ends. Reported-and-tested-by:
-
Ingo Molnar authored
objtool, kprobes/x86: Sync the latest <asm/insn.h> header with tools/objtool/arch/x86/include/asm/insn.h The following commit: ee6a7354: kprobes/x86: Prohibit probing on exception masking instructions Modified <asm/insn.h>, adding the insn_masking_exception() function. Sync the tooling version of the header to it, to fix this warning: Warning: synced file at 'tools/objtool/arch/x86/include/asm/insn.h' differs from latest kernel version at 'arch/x86/include/asm/insn.h' Cc: Peter Zijlstra <peterz@infradead.org> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ricardo Neri <ricardo.neri-calderon@linux.intel.com> Cc: Francis Deslauriers <francis.deslauriers@efficios.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Alexei Starovoitov <ast@kernel.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: "H . Peter Anvin" <hpa@zytor.com> Cc: Yonghong Song <yhs@fb.com> Cc: Borislav Petkov <bp@suse.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: "David S . Miller" <davem@davemloft.net> Signed-off-by:
-
- 13 May, 2018 7 commits
-
-
Alexei Starovoitov authored
Workaround for the sake of BPF compilation which utilizes kernel headers, but clang does not support ASM GOTO and fails the build. Fixes: d0266046 ("x86: Remove FAST_FEATURE_TESTS") Suggested-by:
Alexei Starovoitov <ast@kernel.org> Signed-off-by:
-
Masami Hiramatsu authored
Since MOV SS and POP SS instructions will delay the exceptions until the next instruction is executed, single-stepping on it by uprobes must be prohibited. uprobe already rejects probing on POP SS (0x1f), but allows probing on MOV SS (0x8e and reg == 2). This checks the target instruction and if it is MOV SS or POP SS, returns -ENOTSUPP to reject probing. Signed-off-by:
Masami Hiramatsu <mhiramat@kernel.org> Signed-off-by:
Oleg Nesterov <oleg@redhat.com> Cc: Ricardo Neri <ricardo.neri-calderon@linux.intel.com> Cc: Francis Deslauriers <francis.deslauriers@efficios.com> Cc: Alexei Starovoitov <ast@kernel.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: "H . Peter Anvin" <hpa@zytor.com> Cc: Yonghong Song <yhs@fb.com> Cc: Borislav Petkov <bp@suse.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: "David S . Miller" <davem@davemloft.net> Link: https://lkml.kernel.org/r/152587072544.17316.5950935243917346341.stgit@devbox
-
Masami Hiramatsu authored
Since MOV SS and POP SS instructions will delay the exceptions until the next instruction is executed, single-stepping on it by kprobes must be prohibited. However, kprobes usually executes those instructions directly on trampoline buffer (a.k.a. kprobe-booster), except for the kprobes which has post_handler. Thus if kprobe user probes MOV SS with post_handler, it will do single-stepping on the MOV SS. This means it is safe that if it is used via ftrace or perf/bpf since those don't use the post_handler. Anyway, since the stack switching is a rare case, it is safer just rejecting kprobes on such instructions. Signed-off-by:
-
Tetsuo Handa authored
>From ff82bedd3e12f0d3353282054ae48c3bd8c72012 Mon Sep 17 00:00:00 2001 From: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Date: Wed, 9 May 2018 12:12:39 +0900 Subject: [PATCH v3] x86/kexec: avoid double free_page() upon do_kexec_load() failure. syzbot is reporting crashes after memory allocation failure inside do_kexec_load() [1]. This is because free_transition_pgtable() is called by both init_transition_pgtable() and machine_kexec_cleanup() when memory allocation failed inside init_transition_pgtable(). Regarding 32bit code, machine_kexec_free_page_tables() is called by both machine_kexec_alloc_page_tables() and machine_kexec_cleanup() when memory allocation failed inside machine_kexec_alloc_page_tables(). Fix this by leaving the error handling to machine_kexec_cleanup() (and optionally setting NULL after free_page()). [1] https://syzkaller.appspot.com/bug?id=91e52396168cf2bdd572fe1e1bc0bc645c1c6b40 Fixes: f5deb796 ("x86: kexec: Use one page table in x86_64 machine_kexec") Fixes: 92be3d6b ("kexec/i386: allocate page table pages dynamically") Reported-by:
syzbot <syzbot+d96f60296ef613fe1d69@syzkaller.appspotmail.com> Signed-off-by:
Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by:
Baoquan He <bhe@redhat.com> Cc: thomas.lendacky@amd.com Cc: prudo@linux.vnet.ibm.com Cc: Huang Ying <ying.huang@intel.com> Cc: syzkaller-bugs@googlegroups.com Cc: takahiro.akashi@linaro.org Cc: H. Peter Anvin <hpa@zytor.com> Cc: akpm@linux-foundation.org Cc: dyoung@redhat.com Cc: kirill.shutemov@linux.intel.com Link: https://lkml.kernel.org/r/201805091942.DGG12448.tMFVFSJFQOOLHO@I-love.SAKURA.ne.jp
-
Thomas Gleixner authored
No point to have it at the call sites. Signed-off-by: -
David Wang authored
Centaur CPUs enumerate the cache topology in the same way as Intel CPUs, but the function is unused so for. The Centaur init code also misses to initialize x86_info::max_cores, so the CPU topology can't be described correctly. Initialize x86_info::max_cores and invoke init_cacheinfo() to make CPU and cache topology information available and correct. Signed-off-by:
David Wang <davidwang@zhaoxin.com> Signed-off-by:
-
David Wang authored
There is no point in having the conditional cpu_detect_cache_sizes() call at the callsite of init_intel_cacheinfo(). Move it into init_intel_cacheinfo() and make init_intel_cacheinfo() void. [ tglx: Made the init_intel_cacheinfo() void as the return value was pointless. Adjust changelog accordingly ] Signed-off-by:
-
