DPDK patches and discussions
 help / color / mirror / Atom feed
From: Neil Horman <nhorman@tuxdriver.com>
To: dev@dpdk.org
Subject: [dpdk-dev] [PATCH] eal: fix up bad asm in rte_cpu_get_features
Date: Tue, 18 Mar 2014 16:43:34 -0400	[thread overview]
Message-ID: <1395175414-25232-1-git-send-email-nhorman@tuxdriver.com> (raw)

The recent conversion to build dpdk as a DSO has an error in
rte_cpu_get_features.  When being build with -fpie, %ebx gets clobbered by the
cpuid instruction which is also the pic register.  Therefore the inline asm
tries to save off %ebx, but does so incorrectly.  It starts by loading
params.ebx to "D" which is %edi, but then the first instruction moves %ebx to
%edi, clobbering the input value. Then after the operation is complete, "D"
(%edi) is stored to the local ebx variable, but only after the xchgl instruction
has happened, which means ebx holds only the PIC pointer.  This behavior was
causing strange segfults for me when running the cpuid instruction.

The fix is pretty easy, split the asm into two separate directives, the first
saving ebx, and using it to grab the appropriate cpuid info (and correctly
listing %edi as a clobbered register in the process, and then a subsequent asm
directive preforming the reverse exchange (again, listing %edi as being
clobbered).

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
---
 lib/librte_eal/common/eal_common_cpuflags.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/lib/librte_eal/common/eal_common_cpuflags.c b/lib/librte_eal/common/eal_common_cpuflags.c
index 1ebf78c..2072a0c 100644
--- a/lib/librte_eal/common/eal_common_cpuflags.c
+++ b/lib/librte_eal/common/eal_common_cpuflags.c
@@ -208,16 +208,19 @@ rte_cpu_get_features(struct cpuid_parameters_t params)
 	asm volatile ( 
             "mov %%ebx, %%edi\n"
             "cpuid\n"
-            "xchgl %%ebx, %%edi;\n"
             : "=a" (eax),
-              "=D" (ebx),
+              "=b" (ebx),
               "=c" (ecx),
               "=d" (edx)
             /* input */
             : "a" (params.eax),
-              "D" (params.ebx),
+              "b" (params.ebx),
               "c" (params.ecx),
-              "d" (params.edx));
+              "d" (params.edx)
+	    : "%edi");
+
+	asm volatile ("xchgl %%ebx, %%edi;\n"
+		      : :);
 #endif
 
 	switch (params.return_register) {
-- 
1.8.3.1

             reply	other threads:[~2014-03-18 20:42 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-03-18 20:43 Neil Horman [this message]
2014-03-19 14:48 ` [dpdk-dev] [PATCH v2] " Neil Horman
2014-03-19 15:44   ` H. Peter Anvin
2014-03-20  0:40     ` Neil Horman
2014-03-20  4:22       ` H. Peter Anvin
2014-03-20 11:03         ` Neil Horman
2014-03-20 11:27           ` Neil Horman
2014-03-20 15:20             ` H. Peter Anvin
2014-03-20 11:42 ` [dpdk-dev] [PATCH v3] eal: Fix up assembly for x86_64 " Neil Horman

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=1395175414-25232-1-git-send-email-nhorman@tuxdriver.com \
    --to=nhorman@tuxdriver.com \
    --cc=dev@dpdk.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).