FIX oom and metablk corruption

eBay · Sep 18, 2024 · e7453e4 · e7453e4
1 parent 2a0cf2d
commit e7453e4
Show file tree

Hide file tree

Showing 4 changed files with 9 additions and 6 deletions.
diff --git a/conanfile.py b/conanfile.py
@@ -2,7 +2,7 @@
 
 class HomestoreConan(ConanFile):
     name = "homestore"
-    version = "3.7.1"
+    version = "3.7.2"
 
     homepage = "https://github.corp.ebay.com/SDS/homestore"
     description = "HomeStore"

diff --git a/src/.clang-format b/src/.clang-format
@@ -18,7 +18,6 @@ AlignOperands:   false
 AlignTrailingComments: true
 AllowShortBlocksOnASingleLine: true
 AllowShortIfStatementsOnASingleLine: true
-AllowShortBlocksOnASingleLine: true
 AllowShortCaseLabelsOnASingleLine: false
 # AllowShortFunctionsOnASingleLine: InlineOnly
 # AllowShortLoopsOnASingleLine: false

diff --git a/src/engine/blkstore/blkstore.hpp b/src/engine/blkstore/blkstore.hpp
@@ -54,7 +54,7 @@ VENUM(BlkStoreCacheType, uint8_t, PASS_THRU = 0, WRITEBACK_CACHE = 1, WRITETHRU_
  * be either be discarded or copied into new buffer. This threshold dictates whats the value of (64K - N) upto which
  * it will copy. In other words ((64K - N) <= CACHE_DISCARD_THRESHOLD_SIZE) ? copy : discard
  */
-//#define CACHE_DISCARD_THRESHOLD_SIZE 16384
+// #define CACHE_DISCARD_THRESHOLD_SIZE 16384
 
 class BlkStoreConfig {
 public:
@@ -325,7 +325,7 @@ class BlkStore {
     BlkAllocStatus alloc_contiguous_blk(const uint32_t size, blk_alloc_hints& hints, BlkId* const out_blkid) {
         // Allocate a block from the device manager
         assert(size % m_pagesz == 0);
-        const blk_count_t nblks{static_cast< blk_count_t >(size / m_pagesz)};
+        const uint32_t nblks{static_cast< uint32_t >(size / m_pagesz)};
         hints.is_contiguous = true;
         HS_DBG_ASSERT_LE(nblks, BlkId::max_blks_in_op(), "nblks {} more than max blks {}", nblks,
                          BlkId::max_blks_in_op());
@@ -336,14 +336,17 @@ class BlkStore {
     BlkAllocStatus alloc_blk(const uint32_t size, blk_alloc_hints& hints, std::vector< BlkId >& out_blkid) {
         // Allocate a block from the device manager
         assert(size % m_pagesz == 0);
-        blk_count_t nblks{static_cast< blk_count_t >(size / m_pagesz)};
+        // Using a 16-bit value (blk_count_t) for page counts can lead to overflow issues when managing large storage
+        // devices. With a page size of 4096 bytes, a 16-bit value limits the addressable storage to 256 MB (65,536
+        // pages * 4096 bytes/page).
+        uint32_t nblks{static_cast< uint32_t >(size / m_pagesz)};
         if (nblks <= BlkId::max_blks_in_op()) {
             return (m_vdev.alloc_blk(nblks, hints, out_blkid));
         } else {
             while (nblks != 0) {
                 static thread_local std::vector< BlkId > result_blkid{};
                 result_blkid.clear();
-                const blk_count_t nblks_op{std::min(static_cast< blk_count_t >(BlkId::max_blks_in_op()), nblks)};
+                const uint32_t nblks_op{std::min(static_cast< uint32_t >(BlkId::max_blks_in_op()), nblks)};
                 const auto ret{m_vdev.alloc_blk(nblks_op, hints, result_blkid)};
                 if (ret != BlkAllocStatus::SUCCESS) { return ret; }
                 out_blkid.insert(std::end(out_blkid), std::make_move_iterator(std::begin(result_blkid)),

diff --git a/src/engine/meta/meta_blks_mgr.cpp b/src/engine/meta/meta_blks_mgr.cpp
@@ -1018,6 +1018,7 @@ void MetaBlkMgr::alloc_meta_blk(const uint64_t size, std::vector< BlkId >& bid)
     try {
         const auto ret{m_sb_blk_store->alloc_blk(size, hints, bid)};
         HS_REL_ASSERT_EQ(ret, BlkAllocStatus::SUCCESS);
+        HS_REL_ASSERT_NE(bid.size(), 0);
 #ifndef NDEBUG
         uint64_t debug_size{0};
         for (size_t i{0}; i < bid.size(); ++i) {