From: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
To: Julien Grall <julien.grall@arm.com>
Cc: xen-devel@lists.xenproject.org, ross.lagerwall@citrix.com,
Andrew Cooper <andrew.cooper3@citrix.com>,
sstabellini@kernel.org
Subject: Re: [PATCH v4 04/16] livepatch: Initial ARM64 support.
Date: Mon, 19 Sep 2016 10:33:43 -0400 [thread overview]
Message-ID: <20160919143343.GB9860@localhost.localdomain> (raw)
In-Reply-To: <4979af71-18ca-4c53-b37f-901e7fc4a75b@arm.com>
>
> > void arch_livepatch_revive(void)
> > {
> > + /*
> > + * Nuke the instruction cache. Data cache has been cleaned before in
> > + * arch_livepatch_apply_jmp.
>
> I think you forgot to clean text region from the payload. Without that, you
> may receive a crash if you have a separate cache for data and instruction.
Help me out here please.
Why would we need to call clean_and_invalidate_dcache_va_range on the
payload .text area (the func->new_addr and func->new_size)?
We don't modify that .text area and after this function is done
(arch_livepatch_revive) it would be very first time that code would be called.
Hence there would not be any cache remains at all?
Or did you mean the old_addr (the one we just patched?)
If we are reverting it then we just clear at func->old_addr for one
instruction? We could drop the dcache for the func->new_addr (so new
.text code), to explicitly tell the CPU to not waste cache space for
those instructions? Is that what you meant?
Anyhow did this:
diff --git a/xen/arch/arm/arm64/livepatch.c b/xen/arch/arm/arm64/livepatch.c
index 49eb69b..07f0ce7 100644
--- a/xen/arch/arm/arm64/livepatch.c
+++ b/xen/arch/arm/arm64/livepatch.c
@@ -49,7 +49,10 @@ void arch_livepatch_apply_jmp(struct livepatch_func *func)
for ( i = 0; i < len; i++ )
*(new_ptr + i) = insn;
+ /* There should not be _any_ aliasing using vmap's, but just in case. */
clean_and_invalidate_dcache_va_range(new_ptr, sizeof(*new_ptr) * len);
+ /* And definitly clear the old code. */
+ clean_and_invalidate_dcache_va_range(func->old_addr, sizeof(*new_ptr) * len);
}
void arch_livepatch_revert_jmp(const struct livepatch_func *func)
@@ -68,6 +71,9 @@ void arch_livepatch_revert_jmp(const struct livepatch_func *func)
*(new_ptr + i) = insn;
}
+ /* There should not be _any_ aliasing using vmap's, but just in case. */
+ clean_and_invalidate_dcache_va_range(new_ptr, sizeof(*new_ptr) * len);
+ /* And definitly clear the old code. */
clean_and_invalidate_dcache_va_range(func->old_addr, sizeof(*new_ptr) * len);
}
And added the invalidation of dcache at old_addr (so the function we
patched).
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel
next prev parent reply other threads:[~2016-09-19 14:34 UTC|newest]
Thread overview: 50+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-09-16 16:38 [PATCH v4] Livepatch for ARM 64 and 32 Konrad Rzeszutek Wilk
2016-09-16 16:38 ` [PATCH v4 01/16] arm/x86/common: Add HAS_[ALTERNATIVE|EX_TABLE] Konrad Rzeszutek Wilk
2016-09-19 9:09 ` Jan Beulich
2016-09-19 9:26 ` Julien Grall
2016-09-19 14:04 ` Konrad Rzeszutek Wilk
2016-09-19 14:09 ` Julien Grall
2016-09-19 14:43 ` Konrad Rzeszutek Wilk
2016-09-16 16:38 ` [PATCH v4 02/16] livepatch: Reject payloads with .alternative or .ex_table if support is not built-in Konrad Rzeszutek Wilk
2016-09-16 16:38 ` [PATCH v4 03/16] arm: poison initmem when it is freed Konrad Rzeszutek Wilk
2016-09-19 9:35 ` Julien Grall
2016-09-19 14:19 ` Konrad Rzeszutek Wilk
2016-09-16 16:38 ` [PATCH v4 04/16] livepatch: Initial ARM64 support Konrad Rzeszutek Wilk
2016-09-19 10:26 ` Julien Grall
2016-09-19 14:33 ` Konrad Rzeszutek Wilk [this message]
2016-09-20 9:40 ` Julien Grall
2016-09-16 16:38 ` [PATCH v4 05/16] livepatch: ARM/x86: Check displacement of old_addr and new_addr Konrad Rzeszutek Wilk
2016-09-19 9:19 ` Jan Beulich
2016-09-19 13:12 ` Julien Grall
2016-09-16 16:38 ` [PATCH v4 06/16] livepatch: ARM 32|64: Ignore mapping symbols: $[d, a, x] Konrad Rzeszutek Wilk
2016-09-16 16:38 ` [PATCH v4 07/16] livepatch/arm/x86: Check payload for for unwelcomed symbols Konrad Rzeszutek Wilk
2016-09-19 9:27 ` Jan Beulich
2016-09-19 13:33 ` Julien Grall
2016-09-19 14:11 ` Jan Beulich
2016-09-19 14:13 ` Julien Grall
2016-09-19 14:48 ` Jan Beulich
2016-09-19 17:32 ` Konrad Rzeszutek Wilk
2016-09-20 7:00 ` Jan Beulich
2016-09-20 9:44 ` Julien Grall
2016-09-16 16:38 ` [PATCH v4 08/16] livepatch: Move test-cases to their own sub-directory in test Konrad Rzeszutek Wilk
2016-09-16 16:58 ` Konrad Rzeszutek Wilk
2016-09-16 16:38 ` [PATCH v4 09/16] livepatch: tests: Make them compile under ARM64 Konrad Rzeszutek Wilk
2016-09-19 13:35 ` Julien Grall
2016-09-16 16:38 ` [PATCH v4 10/16] livepatch: x86, ARM, alternative: Expose FEATURE_LIVEPATCH Konrad Rzeszutek Wilk
2016-09-19 13:47 ` Julien Grall
2016-09-16 16:38 ` [PATCH v4 11/16] xen/arm32: Add an helper to invalidate all instruction caches Konrad Rzeszutek Wilk
2016-09-19 14:24 ` Julien Grall
2016-09-16 16:38 ` [PATCH v4 12/16] bug/x86/arm: Align bug_frames sections Konrad Rzeszutek Wilk
2016-09-19 9:29 ` Jan Beulich
2016-09-19 14:34 ` Julien Grall
2016-09-19 14:35 ` Julien Grall
2016-09-19 20:19 ` Konrad Rzeszutek Wilk
2016-09-19 20:26 ` Konrad Rzeszutek Wilk
2016-09-20 9:46 ` Julien Grall
2016-09-16 16:38 ` [PATCH v4 13/16] livepatch: Initial ARM32 support Konrad Rzeszutek Wilk
2016-09-19 14:39 ` Julien Grall
2016-09-16 16:38 ` [PATCH v4 14/16] livepatch, arm[32|64]: Share arch_livepatch_revert_jmp Konrad Rzeszutek Wilk
2016-09-19 14:43 ` Julien Grall
2016-09-16 16:38 ` [PATCH v4 15/16] livepatch: arm[32, 64], x86: NOP test-case Konrad Rzeszutek Wilk
2016-09-16 16:38 ` [PATCH v4 16/16] livepatch: In xen_nop test-case remove the .bss and .data sections Konrad Rzeszutek Wilk
2016-09-19 9:32 ` Jan Beulich
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20160919143343.GB9860@localhost.localdomain \
--to=konrad.wilk@oracle.com \
--cc=andrew.cooper3@citrix.com \
--cc=julien.grall@arm.com \
--cc=ross.lagerwall@citrix.com \
--cc=sstabellini@kernel.org \
--cc=xen-devel@lists.xenproject.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).