[HIVE-20664] Potential ArrayIndexOutOfBoundsException in VectorizedOrcAcidRowBatchReader.findMinMaxKeys

• Type: Sub-task
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 4.0.0-alpha-1
• Component/s: Transactions
• Target Version/s: 4.0.0
• Release Note: n/a
• Assignee: Saurabh Seth
• Reporter: Saurabh Seth
• Created: 01/Oct/18 15:48
• Updated: 17/Nov/22 08:55
• Resolved: 05/Oct/18 16:13

Description

ekoifman, could you please confirm if my understanding is correct and if so, review the fix?

In the method VectorizedOrcAcidRowBatchReader.findMinMaxKeys, the code snippet that identifies the first and last stripe indices in the current split could result in an ArrayIndexOutOfBoundsException if a complete split is within the same stripe:

    for(int i = 0; i < stripes.size(); i++) {
      StripeInformation stripe = stripes.get(i);
      long stripeEnd = stripe.getOffset() + stripe.getLength();
      if(firstStripeIndex == -1 && stripe.getOffset() >= splitStart) {
        firstStripeIndex = i;
      }
      if(lastStripeIndex == -1 && splitEnd <= stripeEnd &&
          stripes.get(firstStripeIndex).getOffset() <= stripe.getOffset() ) {
        //the last condition is for when both splitStart and splitEnd are in
        // the same stripe
        lastStripeIndex = i;
      }
    }

Consider the example where there are 2 stripes - 0-500 and 500-1000 and splitStart is 600 and splitEnd is 800.

In the first iteration of the loop, stripe.getOffset() is 0 and stripeEnd is 500. In this iteration, neither of the if statement conditions will be met and firstSripeIndex as well as lastStripeIndex remain -1.

In the second iteration of the loop stripe.getOffset() is 500, stripeEnd is 1000, The first if statement condition will not be met in this case because stripe's offset (500) is not greater than or equal to the splitStart (600). However, in the second if statement, splitEnd (800) is <= stripeEnd(1000) and it will try to compute the last condition stripes.get(firstStripeIndex).getOffset() <= stripe.getOffset(). This will throw an ArrayIndexOutOfBoundsException because firstStripeIndex is still -1.

I'm not sure if this scenario is possible at all, hence logging this as a low priority issue. Perhaps block based split generation using BISplitStrategy could trigger this?

Issue Links

• is related to: HIVE-20694 Additional unit tests for VectorizedOrcAcidRowBatchReader min max key evaluation