In VACUUM FULL, an interrupt after the initial transaction has been recorded
as committed can cause postmaster to restart with the following error message:
PANIC: cannot abort transaction NNNN, it was already committed
This problem has been reported many times.
In lazy VACUUM, an interrupt after the table has been truncated by
lazy_truncate_heap causes other backends' relcache to still point to the
removed pages; this can cause future INSERT and UPDATE queries to error out
with the following error message:
could not read block XX of relation 1663/NNN/MMMM: read only 0 of 8192 bytes
The window to this race condition is extremely narrow, but it has been seen in
the wild involving a cancelled autovacuum process.
The solution for both problems is to inhibit interrupts in both operations
until after the respective transactions have been committed. It's not a
complete solution, because the transaction could theoretically be aborted by
some other error, but at least fixes the most common causes of both problems.
In PLy_output(), when the elog() call in the TRY branch throws an exception
(this can happen when a statement timeout kicks in, for example), the
PyErr_SetString() call in the CATCH branch can cause a segfault, because the
Py_XDECREF(so) call before it releases memory that is still used by the sv
variable that PyErr_SetString() uses as argument, because sv points into
memory owned by so.
Backpatched back to 8.0, where this code was introduced.
I also threw in a couple of volatile declarations for variables that are used
before and after the TRY. I don't think they caused the crash that I
observed, but they could become issues.
The original coding ensured nbuckets and nbatch didn't exceed INT_MAX,
which while not insane on its own terms did nothing to protect subsequent
code like "palloc(nbatch * sizeof(BufFile *))". Since enormous join size
estimates might well be planner error rather than reality, it seems best
to constrain the initial sizes to be not more than work_mem/sizeof(pointer),
thus ensuring the allocated arrays don't exceed work_mem. We will allow
nbatch to get bigger than that during subsequent ExecHashIncreaseNumBatches
calls, but we should still guard against integer overflow in those palloc
requests. Per bug #5145 from Bernt Marius Johnsen.
Although the given test case only seems to fail back to 8.2, previous
releases have variants of this issue, so patch all supported branches.
that it's called within an AfterTriggerBeginQuery/AfterTriggerEndQuery pair.
The RI cascade triggers suppress that overhead on the assumption that they
are always run non-deferred, so it's possible to violate the condition if
someone mistakenly changes pg_trigger to mark such a trigger deferred.
We don't really care about supporting that, but throwing an error instead
of crashing seems desirable. Per report from Marcelo Costa.
pam_message array contains exactly one PAM_PROMPT_ECHO_OFF message.
Instead, deal with however many messages there are, and don't throw error
for PAM_ERROR_MSG and PAM_TEXT_INFO messages. This logic is borrowed from
openssh 5.2p1, which hopefully has seen more real-world PAM usage than we
have. Per bug #5121 from Ryan Douglas, which turned out to be caused by
the conv_proc being called with zero messages. Apparently that is normal
behavior given the combination of Linux pam_krb5 with MS Active Directory
as the domain controller.
Patch all the way back, since this code has been essentially untouched
since 7.4. (Surprising we've not heard complaints before.)
in CREATE OR REPLACE FUNCTION. The original code would update pg_shdepend
as if a new function was being created, even if it wasn't, with two bad
consequences: pg_shdepend might record the wrong owner for the function,
and any dependencies for roles mentioned in the function's ACL would be lost.
The fix is very easy: just don't touch pg_shdepend at all when doing a
function replacement.
Also update the CREATE FUNCTION reference page, which never explained
exactly what changes and doesn't change in a function replacement.
In passing, fix the CREATE VIEW reference page similarly; there's no
code bug there, but the docs didn't say what happens.
possibility of shared-inval messages causing a relcache flush while it tries
to fill in missing data in preloaded relcache entries. There are actually
two distinct failure modes here:
1. The flush could delete the next-to-be-processed cache entry, causing
the subsequent hash_seq_search calls to go off into the weeds. This is
the problem reported by Michael Brown, and I believe it also accounts
for bug #5074. The simplest fix is to restart the hashtable scan after
we've read any new data from the catalogs. It appears that pre-8.4
branches have not suffered from this failure, because by chance there were
no other catalogs sharing the same hash chains with the catalogs that
RelationCacheInitializePhase2 had work to do for. However that's obviously
pretty fragile, and it seems possible that derivative versions with
additional system catalogs might be vulnerable, so I'm back-patching this
part of the fix anyway.
2. The flush could delete the *current* cache entry, in which case the
pointer to the newly-loaded data would end up being stored into an
already-deleted Relation struct. As long as it was still deleted, the only
consequence would be some leaked space in CacheMemoryContext. But it seems
possible that the Relation struct could already have been recycled, in
which case this represents a hard-to-reproduce clobber of cached data
structures, with unforeseeable consequences. The fix here is to pin the
entry while we work on it.
In passing, also change RelationCacheInitializePhase2 to Assert that
formrdesc() set up the relation's cached TupleDesc (rd_att) with the
correct type OID and hasoids values. This is more appropriate than
silently updating the values, because the original tupdesc might already
have been copied into the catcache. However this part of the patch is
not in HEAD because it fails due to some questionable recent changes in
formrdesc :-(. That will be cleaned up in a subsequent patch.
It seems the flex developers have decided to change yyleng from int to size_t.
This has already happened in the latest release of OS X, and will start
happening elsewhere once the next release of flex appears. Rather than trying
to divine how it's declared in any particular build, let's just remove the one
existing not-very-necessary external usage.
Back-patch to all supported branches; not so much because users in the field
are likely to care about building old branches with cutting-edge flex, as
to keep OSX-based buildfarm members from having problems with old branches.
functions.
This extends the previous patch that forbade SETting these variables inside
security-definer functions. RESET is equally a security hole, since it
would allow regaining privileges of the caller; furthermore it can trigger
Assert failures and perhaps other internal errors, since the code is not
expecting these variables to change in such contexts. The previous patch
did not cover this case because assign hooks don't really have enough
information, so move the responsibility for preventing this into guc.c.
Problem discovered by Heikki Linnakangas.
Security: no CVE assigned yet, extends CVE-2007-6600
(could happen if either postgresql.conf or postmaster.opts is empty).
It's been broken since the C version was written for 8.0, so patch
all the way back.
initdb's copy of the function is broken in the same way, but it's
less important there since the input files should never be empty.
Patch that in HEAD only, and also fix some cosmetic differences that
crept into that copy of the function.
Per report from Corry Haines and Jeff Davis.
#include the version of history.h that is in the same directory as the
readline.h we are using. This avoids problems in some scenarios where both
readline and editline are installed. Report and patch by Zdenek Kotala.
when we reach the post-COPY "pump it dry" error recovery code that was added
2006-11-24. Per a report from Neil Best, there is at least one code path
in which this occurs, leading to an infinite loop in code that's supposed
to be making it more robust not less so. A reasonable response seems to be
to call PQputCopyEnd() again, so let's try that.
Back-patch to all versions that contain the cleanup loop.
we should ignore NULL array entries, not non-NULL ones. This had the
effect of disabling commit_delay, and could have caused a crash in the
rare race condition the patch was intended to fix.
Bug report and diagnosis by Jeff Janes, in bug #4952.
last pair of parameter name/value strings, even when there are MAXPARAMS
of them. Aboriginal bug in contrib/xml2, noted while studying bug #4912
(though I'm not sure whether there's something else involved in that
report).
This might be thought a security issue, since it's a potential backend
crash; but considering that untrustworthy users shouldn't be allowed
to get their hands on xslt_process() anyway, it's probably not worth
getting excited about.
a number of other geometric operators also depend on. It miscalculated the
slope of the perpendicular to the given line segment anytime that slope was
other than 0, infinite, or +/-1. In some cases the error would be masked
because the true closest point on the line segment was one of its endpoints
rather than the intersection point, but in other cases it could give an
arbitrarily bad answer. Per bug #4872 from Nick Roosevelt.
Bug goes clear back to Berkeley days, so patch all supported branches.
Make a couple of cosmetic adjustments while at it.
eg Japan. Report and fix by Itagaki Takahiro. Also fix CASHDEBUG printout
format for branches with 64-bit money type, and some minor comment cleanup.
Back-patch to 7.4, because it's broken all the way back.
there's no analyzable attributes or indexes. We also used to report 0 live
and dead tuples for such tables, which messed with autovacuum threshold
calculations.
This fixes bug #4812 reported by George Su. Backpatch back to 8.1.
as per my recent proposal. release.sgml itself is now just a stub that should
change rarely; ideally, only once per major release to add a new include line.
Most editing work will occur in the release-N.N.sgml files. To update a back
branch for a minor release, just copy the appropriate release-N.N.sgml
file(s) into the back branch.
This commit doesn't change the end-product documentation at all, only the
source layout. However, it makes it easy to start omitting ancient information
from newer branches' documentation, should we ever decide to do that.
part that rounds up to exactly 1.0 second. The previous coding rejected input
like "00:12:57.9999999999999999999999999999", with the exact number of nines
needed to cause failure varying depending on float-timestamp option and
possibly on platform. Obviously this should round up to the next integral
second, if we don't have enough precision to distinguish the value from that.
Per bug #4789 from Robert Kruus.
In passing, fix a missed check for fractional seconds in one copy of the
"is it greater than 24:00:00" code.
Broken all the way back, so patch all the way back.
aggregate function. By definition, such a sub-SELECT cannot reference any
variables of query levels between itself and the aggregate's semantic level
(else the aggregate would've been assigned to that lower level instead).
So the correct, most efficient implementation is to treat the sub-SELECT as
being a sub-select of that outer query level, not the level the aggregate
syntactically appears in. Not doing so also confuses the heck out of our
parameter-passing logic, as illustrated in bug report from Daniel Grace.
Fortunately, we were already copying the whole Aggref expression up to the
outer query level, so all that's needed is to delay SS_process_sublinks
processing of the sub-SELECT until control returns to the outer level.
This has been broken since we introduced spec-compliant treatment of
outer aggregates in 7.4; so patch all the way back.
interval_eq() considers equal. I'm not sure how that fundamental requirement
escaped us through multiple revisions of this hash function, but there it is;
it's been wrong since interval_hash was first written for PG 7.1.
Per bug #4748 from Roman Kononov.
Backpatch to all supported releases.
This patch changes the contents of hash indexes for interval columns. That's
no particular problem for PG 8.4, since we've broken on-disk compatibility
of hash indexes already; but it will require a migration warning note in
the next minor releases of all existing branches: "if you have any hash
indexes on columns of type interval, REINDEX them after updating".
temporary tables of other sessions; that is unsafe because of the way our
buffer management works. Per report from Stuart Bishop.
This is redundant with the bufmgr.c checks in HEAD, but not at all redundant
in the back branches.
at the same instant as a new backend is spawned. Since CountActiveBackends()
doesn't hold ProcArrayLock, it needs to be prepared for the case that a
pointer at the end of the proc array is still NULL even though numProcs says
it should be valid, since it doesn't hold ProcArrayLock. Backpatch to 8.1.
8.0 and earlier had this right, but it was broken in the split of PGPROC and
sinval shared memory arrays.
Per report and proposal by Marko Kreen.
not global variables of anonymous enum types. This didn't actually hurt
much because most linkers will just merge the duplicated definitions ...
but some will complain. Per bug #4731 from Ceriel Jacobs.
Backpatch to 8.1 --- the declarations don't exist before that.
format codes are misapplied to a numeric argument. (The code still produces
a pretty bogus error message in such cases, but I'll settle for stopping the
crash for now.) Per bug #4700 from Sergey Burladyan.
Problem exists in all supported branches, so patch all the way back.
In HEAD, also clean up some ugly coding in the nearby cache management
code.
Alvaro Herrera