| 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
| ||
|
Date: Mon, 24 Oct 2022 18:04:08 -0400
From: Steven Rostedt <rostedt@...dmis.org>
To: Linus Torvalds <torvalds@...ux-foundation.org>
Cc: LKML <linux-kernel@...r.kernel.org>,
Peter Zijlstra <peterz@...radead.org>,
Kees Cook <keescook@...omium.org>,
Dave Hansen <dave.hansen@...el.com>,
Sean Christopherson <seanjc@...gle.com>
Subject: Re: [PATCH] x86/mm: Do not verify W^X at boot up
On Mon, 24 Oct 2022 12:08:49 -0700
Linus Torvalds <torvalds@...ux-foundation.org> wrote:
> And then all these special cases beget other special cases.
Well, I was able to get it to work without these special cases, but it
caused a bit of another kind of special cases to get poking_init() into
mm_init().
To get poking_init() working in mm_init() I had to pull in:
proc_caches_init(), as poking_init() uses some fork code that requires its
caches to be initialized.
Then dup_mm() is called, which uses maple tree code, which required
maple_tree_init() to be there too. (I pulled in radix_tree_init() just to
be consistent). But maple tree code calls kmem_cache_alloc_bulk() which
specifically states:
/* Note that interrupts must be enabled when calling this function. */
and lockdep confirmed it.
So I did some hacking in the maple_tree.c to make that work.
And finally, dup_mm() calls dup_mmap() that calls flush_tlb_mm() for the
old mm, but since this is early boot up, there's really no need for that. I
added some hacks to avoid that.
Thus, I guess you get to choose your poison. Either we have special ftrace
cases in x86 that beget other special cases to keep it working, or we make
text_poke() work early by moving poking_init() into mm_init() and then
creating more generic special cases that beget other special cases (and I
have no idea if this works on other architectures, which could beget more
special cases).
Your call.
-- Steve
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 5cadcea035e0..e240351e0bc1 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -1681,11 +1681,6 @@ void __ref text_poke_queue(void *addr, const void *opcode, size_t len, const voi
{
struct text_poke_loc *tp;
- if (unlikely(system_state == SYSTEM_BOOTING)) {
- text_poke_early(addr, opcode, len);
- return;
- }
-
text_poke_flush(addr);
tp = &tp_vec[tp_vec_nr++];
@@ -1707,11 +1702,6 @@ void __ref text_poke_bp(void *addr, const void *opcode, size_t len, const void *
{
struct text_poke_loc tp;
- if (unlikely(system_state == SYSTEM_BOOTING)) {
- text_poke_early(addr, opcode, len);
- return;
- }
-
text_poke_loc_init(&tp, addr, opcode, len, emulate);
text_poke_bp_batch(&tp, 1);
}
diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c
index bd165004776d..43628b8480fa 100644
--- a/arch/x86/kernel/ftrace.c
+++ b/arch/x86/kernel/ftrace.c
@@ -415,8 +415,7 @@ create_trampoline(struct ftrace_ops *ops, unsigned int *tramp_size)
set_vm_flush_reset_perms(trampoline);
- if (likely(system_state != SYSTEM_BOOTING))
- set_memory_ro((unsigned long)trampoline, npages);
+ set_memory_ro((unsigned long)trampoline, npages);
set_memory_x((unsigned long)trampoline, npages);
return (unsigned long)trampoline;
fail:
diff --git a/init/main.c b/init/main.c
index aa21add5f7c5..e5f4ae2d4cca 100644
--- a/init/main.c
+++ b/init/main.c
@@ -860,6 +860,10 @@ static void __init mm_init(void)
/* Should be run after espfix64 is set up. */
pti_init();
kmsan_init_runtime();
+ proc_caches_init();
+ radix_tree_init();
+ maple_tree_init();
+ poking_init();
}
#ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
@@ -1011,8 +1015,6 @@ asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
if (WARN(!irqs_disabled(),
"Interrupts were enabled *very* early, fixing it\n"))
local_irq_disable();
- radix_tree_init();
- maple_tree_init();
/*
* Set up housekeeping before setting up workqueues to allow the unbound
@@ -1117,7 +1119,6 @@ asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
thread_stack_cache_init();
cred_init();
fork_init();
- proc_caches_init();
uts_ns_init();
key_init();
security_init();
@@ -1134,7 +1135,6 @@ asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
taskstats_init_early();
delayacct_init();
- poking_init();
check_bugs();
acpi_subsystem_init();
diff --git a/kernel/fork.c b/kernel/fork.c
index 08969f5aa38d..e24fb3ddcf9f 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -702,7 +702,8 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
mas_destroy(&mas);
out:
mmap_write_unlock(mm);
- flush_tlb_mm(oldmm);
+ if (likely(!early_boot_irqs_disabled))
+ flush_tlb_mm(oldmm);
mmap_write_unlock(oldmm);
dup_userfaultfd_complete(&uf);
fail_uprobe_end:
diff --git a/lib/maple_tree.c b/lib/maple_tree.c
index e1743803c851..6fc72ca62c7d 100644
--- a/lib/maple_tree.c
+++ b/lib/maple_tree.c
@@ -1253,7 +1253,12 @@ static inline void mas_alloc_nodes(struct ma_state *mas, gfp_t gfp)
}
max_req = min(requested, max_req);
- count = mt_alloc_bulk(gfp, max_req, slots);
+ if (unlikely(early_boot_irqs_disabled)) {
+ slots[0] = mt_alloc_one(gfp | GFP_ATOMIC);
+ count = slots[0] ? 1 : 0;
+ } else {
+ count = mt_alloc_bulk(gfp, max_req, slots);
+ }
if (!count)
goto nomem_bulk;
Powered by blists - more mailing lists