Closed
Bug 1274313
Opened 9 years ago
Closed 9 years ago
NSS trunk fails when both NSS_ENABLE_TLS_1_3=1 and NSS_NO_PKCS11_BYPASS=1 are defined
Categories
(NSS :: Libraries, defect)
Tracking
(Not tracked)
RESOLVED
FIXED
3.25
People
(Reporter: KaiE, Assigned: ttaubert)
Details
Attachments
(1 file)
1.12 KB,
patch
|
ekr
:
review+
|
Details | Diff | Splinter Review |
When building NSS trunk with both
NSS_ENABLE_TLS_1_3=1
NSS_NO_PKCS11_BYPASS=1
the ssl_gtest crashes.
It works fine when only one of them is defined (regardless which one).
I tried to investigate the crash, but my initial attempts have failed.
I see results which are difficult to explain, and might suggest a memory error.
I've used the following patch for tracing:
https://paste.fedoraproject.org/368505/
I've got these results prior to the crash:
https://paste.fedoraproject.org/368507/
Compare the first line containing ==== with the last one:
===== in ssl3_InitState calling PR_INIT_CLIST, ss:0xbff5f0 &ss->ssl3:0xbff9e8 &cipherSpecs:0xbffd60
===== in SSLInt_CountTls13CipherSpecs, ss:0xbff5f0 &ss->ssl3:0xbff9e8 &cipherSpecs:0xc00090
How could &ss->ssl3.hs.cipherSpecs possibly change,
if cipherSpecs is a member of hs, which is a member of ssl3,
and both ss and &ss->ssl3 haven't changed?
Assignee | ||
Updated•9 years ago
|
Assignee: nobody → ttaubert
Status: NEW → ASSIGNED
Assignee | ||
Comment 1•9 years ago
|
||
(In reply to Kai Engert (:kaie) from comment #0)
> How could &ss->ssl3.hs.cipherSpecs possibly change,
> if cipherSpecs is a member of hs, which is a member of ssl3,
> and both ss and &ss->ssl3 haven't changed?
The SSL3HandshakeState struct looks different with NO_PKCS11_BYPASS=1:
https://hg.mozilla.org/projects/nss/annotate/c4e46ec959d3/lib/ssl/sslimpl.h#l868
libssl is properly built with the smaller version of the struct, ssl_gtests however doesn't know about that define, includes sslimpl.h, and assumes the bigger version of the struct.
Assignee | ||
Comment 2•9 years ago
|
||
Attachment #8759974 -
Flags: review?(martin.thomson)
Attachment #8759974 -
Flags: review?(ekr)
Assignee | ||
Updated•9 years ago
|
Attachment #8759974 -
Flags: review?(franziskuskiefer)
Comment 3•9 years ago
|
||
Comment on attachment 8759974 [details] [diff] [review]
0001-Bug-1274313-Ensure-NSS_NO_PKCS11_BYPASS-1-is-propaga.patch
Review of attachment 8759974 [details] [diff] [review]:
-----------------------------------------------------------------
LGTM. You should be able to commit it without two more reviews :)
::: coreconf/config.mk
@@ +211,5 @@
>
> +# Allow disabling PKCS11 bypass.
> +ifdef NSS_NO_PKCS11_BYPASS
> +DEFINES += -DNO_PKCS11_BYPASS
> +endif
LGTM
Attachment #8759974 -
Flags: review?(martin.thomson)
Attachment #8759974 -
Flags: review?(franziskuskiefer)
Attachment #8759974 -
Flags: review?(ekr)
Attachment #8759974 -
Flags: review+
Assignee | ||
Comment 4•9 years ago
|
||
(In reply to Eric Rescorla (:ekr) from comment #3)
> LGTM. You should be able to commit it without two more reviews :)
Yes, that was the plan :) Thanks!
Assignee | ||
Comment 5•9 years ago
|
||
Status: ASSIGNED → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
Target Milestone: --- → 3.25
You need to log in
before you can comment on or make changes to this bug.
Description
•