Red Hat Bugzilla – Bug 59102
rpm returns: DB_VERIFY_BAD: Database verification failed
Last modified: 2016-09-29 01:07:30 EDT
Description of Problem: I was trying to upgrade to XFree86 4.2.0 to try to pinpoint an error that exists with my graphics card and X. So I grabbed XFree86 from Rawhide along with gcc-3.1, libstdc++, cpp and rpm (and DB4 for RPM). All packages except Glide and XFree86-4.2.0-3.2-i386.rpm installed correctly (this includes XFree86-devel, libs, tools, xfs and so one). However Glide and XFree86 returned errors similiar to these: # rpm -Uvh Glide3-20010520-10.i386.rpm --nodeps Preparing... ########################################### [100%] 1:Glide3 ########################################### [100%] rpmdb: Suspiciously high nelem of 4294967294 on page 0 error: db4 error(-30979) from db->verify: DB_VERIFY_BAD: Database verification failed Version-Release number of selected component (if applicable): RPM version 4.0.4-0.22 XFree86 version 4.2.0 Glide version 20010520-10-3.2 How Reproducible: Every time Steps to Reproduce: 1. Install the RPM 2. 3. Actual Results: An error code from RPM Expected Results: No error code Additional Information: I marked this as high since the verification fails meaning that the database might be corrupted. An rpm -V XFree86 returns som .......T errors but nothing else.
I have also been observing this behaviour lately, however have not installed the packages that you described above. This leads me to believe it's not an issue with those particular packages, but with rpm itself. For me it has happened with a seemingly random selection of installs - when it breaks it breaks for every package until I do the workaround as listed below. A workaround, albeit temporary, is to do an rpm --rebuilddb This seems to get rid of the errors for a while however they eventually return. This problem was observed with a handful of releases prior to 4.0.4-20 as well, so I'm guessing it is an unresolved bug...
Yup, an unresolved bug that cannot be reproduced ATM, but may be related to locking on SMP as well as ^C handling. Reopen this bug if you have a reproducible test case; otherwise the fix is do "rpm --rebuilddb" until I can reprodcue the problem.
I've actually logged a new bug for what may be the same problem; feel free to mark it as a duplicate.
Search can't find it, maybe I didn't log it. Anyway, more info: it seems to be the "Conflictnames" file, and I'm attaching a db_dump of mine (it's not too big). Looks like a db bug (element count being reduced too many times when elements removed) ?
Quick and dirty fix: db_dump Conflictname >/tmp/Conflictname.db_dump [edit latter file, changing h_nelem to 1] db_load Conflictname <!$ A subsequent db_dump shows that the element count has been corrected (to 59) when loading it. However it breaks again if I install and then remove the same packages I was working with when the fault showed up. Log attached in a mo.
Created attachment 44358 [details] Log of actions to reproduce the error we are encountering (the ls at the end shows packages from rawhide-20020201 which trigger the problem)
Hmmm I have a patch in rpm-4.0.3 for a page refcount problem on DB_HASH's, will have to check to see that the patch made it into db-4.0.14.
Let me know if I can provide any more info (the Conflictname table is quite small so I could attach a copy of the db_dump, for example).
Another package that triggered the problem, if it's any help: [root@pikachu tmp]# rpm -e nautilus-mozilla rpmdb: Suspiciously high nelem of 4294967295 on page 0 error: db4 error(-30979) from db->verify: DB_VERIFY_BAD: Database verification failed rpmdb: Suspiciously high nelem of 4294967295 on page 0 error: db4 error(-30979) from db->verify: DB_VERIFY_BAD: Database verification failed
(that last one screwed up the Name table as well :o(
*** This bug has been marked as a duplicate of 50679 ***