Closed Bug 678440 Opened 11 years ago Closed 11 years ago
.0a1 Crash [@ ns NSSSocket Info::Ensure Doc Shell Dependent Stuff Known()]
User Agent: Mozilla/5.0 (Windows NT 6.1; rv:8.0a1) Gecko/20110811 Firefox/8.0a1 Build ID: 20110811152156 Steps to reproduce: STR 1. Downloaded https://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-central-win32/1313101316/firefox-8.0a1.en-US.win32.zip 2. Extract and ran firefox -P with new profile 3. Installed http://www.echofon.com/twitter/firefox/Echofon.xpi and restart 4. Right click on Echofon toolbar, preferences, advanced, check Use SSL for all requests (This is important), and authorize Echofon 5. when authroized, firefox crash Note: firefox won't crash if uncheck "Use SSL for all requests". Actual results: Crash [@ xul.dll@0x9cd449 ] 0 xul.dll xul.dll@0x9cd449 1 xul.dll xul.dll@0x9d830c 2 ssl3.dll ssl3.dll@0x6d4f 3 ssl3.dll ssl3.dll@0x9563 4 ssl3.dll ssl3.dll@0x96cd 5 ssl3.dll ssl3.dll@0x9b79 6 ssl3.dll ssl3.dll@0x9eb5 7 ssl3.dll ssl3.dll@0xb022 8 ssl3.dll ssl3.dll@0xfb0d 9 ssl3.dll ssl3.dll@0x10920 10 ssl3.dll ssl3.dll@0x109ad 11 ssl3.dll ssl3.dll@0x135f9 12 xul.dll xul.dll@0x42f545 13 nspr4.dll nspr4.dll@0x220f 14 nspr4.dll nspr4.dll@0x1d2c 15 mozcrt19.dll mozcrt19.dll@0x2c27 16 mozcrt19.dll mozcrt19.dll@0xd297 17 mozcrt19.dll mozcrt19.dll@0x2cb5 18 kernel32.dll BaseThreadInitThunk 19 ntdll.dll __RtlUserThreadStart 20 kernel32.dll WerpReportFaultInternal 21 kernel32.dll WerpReportFaultInternal bp-29258595-2ce0-4bf0-9760-b35632110811 bp-ab4a6d16-8704-4666-99e5-286312110811 bp-d8ffd816-a90e-475f-b6c1-676852110811 Regression Good http://hg.mozilla.org/mozilla-central/rev/42f7ed136034 Bad http://hg.mozilla.org/mozilla-central/rev/10915aa17365 Mounir Lamouri — Merging mozilla-inbound into mozilla-central. m-i Good http://hg.mozilla.org/integration/mozilla-inbound/rev/dd0630c44b2d http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-inbound-win32/1312999277/ Bad http://hg.mozilla.org/integration/mozilla-inbound/rev/be91fb29d950 http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-inbound-win32/1313002780/
Severity: normal → critical
Crash Signature: [@ xul.dll@0x9cd449 ] [@ xul.dll@0x9cdc17 ] [@ xul.dll@0x9cdc0d ]
Confirmed on nightly Nightly linux build. http://hg.mozilla.org/mozilla-central/rev/f262c389193e Mozilla/5.0 (X11; Linux i686; rv:8.0a1) Gecko/20110812 Firefox/8.0a1 ID:20110812030744 bp-7cb93135-30a5-4729-8de1-c2d5d2110812 bp-e8f478cf-a441-4574-821c-728772110812
And also crashes on http://hg.mozilla.org/mozilla-central/rev/f262c389193e Mozilla/5.0 (Windows NT 6.1; WOW64; rv:8.0a1) Gecko/20110812 Firefox/8.0a1 ID:20110812030744 bp-986a672b-0872-4dd9-b048-5d1de2110812 bp-0e2e0952-9778-4369-843d-f4e9a2110812
Status: UNCONFIRMED → NEW
Crash Signature: [@ xul.dll@0x9cd449 ] [@ xul.dll@0x9cdc17 ] [@ xul.dll@0x9cdc0d ] → [@ nsNSSSocketInfo::EnsureDocShellDependentStuffKnown()]
Ever confirmed: true
OS: Windows 7 → All
Summary: Firefox 8.0a1 Crash [@ xul.dll@0x9cd449 ] → Firefox 8.0a1 Crash [@ nsNSSSocketInfo::EnsureDocShellDependentStuffKnown()]
triggered by be91fb29d950 Luke Wagner — Bug 674597 - abort if attempting to create an xpcom proxy for wrapped JS (r=bsmedberg)
Component: General → Security: PSM
Product: Firefox → Core
QA Contact: general → psm
me too Mozilla/5.0 (Windows NT 5.1; rv:8.0a1) Gecko/20110812 Firefox/8.0a1 ID:20110812030744
Well, it looks like we have a binary component attempting to create an xpcom proxy for wrapped JS. This is an invalid/unsafe operation.
To be clear: that's all bug 674597 does; check and abort.
Can reproduce, the offending NS_GetProxyForObject is http://hg.mozilla.org/mozilla-central/file/cec797d85529/security/manager/ssl/src/nsNSSIOLayer.cpp#l394; Echofon must be implementing a nsIDocShellTreeItem.
This fixes the assert for me. I just did a simple conversion to runnable like bug 674571. Kai: before bug 674597 and with this patch, the "rootItem do_QI is null" assert fires (this is with Echofon installed); can I remove it?
Assignee: nobody → luke
Status: NEW → ASSIGNED
Comment on attachment 552804 [details] [diff] [review] fix Wrong.
Kai, there was a peculiar thing (we think a bug) where NS_ProxyRelease was being called in secureUI right before secureUI was used again. I just wanted to call your attention to this so you could double check.
Luke, I found another problem similar to this in the same code path: we will sometimes try to create an XPCOM proxy for an nsIPrompt that is implemented by JS. I am working on a fix that this, that, and at least one other related bug with a single runnable. I would rather use that solution than this one.
Brian, thanks a lot for your help on this. I have two patches converting two more sites where proxies are used into runnables (bug 678440 and bug 678547), one of which is an nsIPrompt, so hopefully this helps. I'm now actively investigating this bug.
Oops, I have my bugs confused; this is bug 678440 and I am investigating bug 678706. But bug 678547 does have the nsIPrompt patch.
EnsureDocShellDependentStuffKnown is called (through GetPreviousCert) on the critical path of certificate chain validation in the SSL handshake. AFAICT, adding a sync dispatch to it will cause certificate path validation--and all network I/O--to wait for the main thread's event queue to drain. I do not know how much of a performance impact that would have but I think I have an alternative solution with better performance characteristics. Will post soon.
IIUC, this wouldn't be *adding* a sync dispatch, but replacing the (slower) one inherent in proxies. Actually, this patch replaces three proxy dispatches (each doing a round-trip sync). You know the code better so I'm happy to let you do the patch though.
bsmith has some patches to kill most proxies in ssl (bsmedberg also has a patch under review to remove *all* xpcom/proxies). However, neither of these will be landing in the next day or two, so bsmith and I agreed to just back out the offending patch and dependent asserts. This is just a backout, so no r=, just approval to land on just-branched aurora.
Comment on attachment 553609 [details] [diff] [review] revert Approved for mozilla-aurora
Attachment #553609 - Flags: approval-mozilla-aurora? → approval-mozilla-aurora+
Crashes have dropped to 0, marking fixed.
Status: ASSIGNED → RESOLVED
Closed: 11 years ago
Resolution: --- → FIXED
Build identifier: Mozilla/5.0 (Windows NT 6.1; rv:8.0) Gecko/20100101 Firefox/8.0 I tried to install the plugin from the link in bug description but plugin can not be installed because it is incompatible with Firefox 8.0. Is there another version of plugin that is compatible with Firefox 8? Thank you!
Mihaela, I notice that Echofon 2.2.4 beta 1 compatible with Firefox 8 http://www.echofon.com/twitter/firefox/Echofon_beta.xpi
Duplicate of this bug: 679035
Duplicate of this bug: 679030
Mozilla/5.0 (Windows NT 5.1; rv:10.0a1) Gecko/20111006 Firefox/10.0a1 Mozilla/5.0 (X11; Linux i686; rv:10.0a1) Gecko/20111006 Firefox/10.0a1 Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:10.0a1) Gecko/20111006 Firefox/10.0a1 Mozilla/5.0 (Windows NT 6.1; rv:10.0a1) Gecko/20111006 Firefox/10.0a1 Mozilla/5.0 (X11; Linux i686; rv:8.0) Gecko/20100101 Firefox/8.0 Mozilla/5.0 (Windows NT 5.1; rv:8.0) Gecko/20100101 Firefox/8.0 Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:8.0) Gecko/20100101 Firefox/8.0 Mozilla/5.0 (Windows NT 6.1; rv:8.0) Gecko/20100101 Firefox/8.0 Verified as fixed on the above builds: Firefox doesn't crash after authorizing or submiting tweet. Still, no data is displayed on nightlies in either of the Echofon's tabs, but I will log another bug for that.
Status: RESOLVED → VERIFIED
Crash Signature: [@ nsNSSSocketInfo::EnsureDocShellDependentStuffKnown()] → [@ nsNSSSocketInfo::EnsureDocShellDependentStuffKnown()] [@ nsHandleSSLError ] [@ RtlDeleteCriticalSection] [@ nsHttpConnection::Close(unsigned int)] [@ nsHttpConnection::EndIdleMonitoring()] [@ displayAlert ] [@ PK11PasswordPrompt(PK11SlotInfoStr* …
Whiteboard: [verified-beta] → [verified-beta] [qa!]
You need to log in before you can comment on or make changes to this bug.