snapuserd: I/O requests which are not block aligned.
If the I/O request is for a sector which is not block
aligned, then we will not find the mapping directly
to a COW op as each op is block aligned. Thus, we
handle them by chopping the I/O request and processing
the unaligned data. Furthermore, if the request
doesn't map to any of the COW ops, then route the
I/O to the base device.
Bug: 196929997
Test: Full OTA on CF - during boot up, we get an I/O request
to read superblock on system partition which is not block aligned.
Signed-off-by: Akilesh Kailash <akailash@google.com>
Change-Id: Ie40633c0c6bc0c87b681c051c74c0ac787e34d9c
diff --git a/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_core.h b/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_core.h
index 38cfef2..d561952 100644
--- a/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_core.h
+++ b/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_core.h
@@ -167,6 +167,9 @@
bool ReadFromSourceDevice(const CowOperation* cow_op);
bool ReadAlignedSector(sector_t sector, size_t sz, bool header_response);
+ bool ReadUnalignedSector(sector_t sector, size_t size);
+ int ReadUnalignedSector(sector_t sector, size_t size,
+ std::vector<std::pair<sector_t, const CowOperation*>>::iterator& it);
bool RespondIOError(bool header_response);
// Processing COW operations
diff --git a/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_dm_user.cpp b/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_dm_user.cpp
index 9209944..bfbacf9 100644
--- a/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_dm_user.cpp
+++ b/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_dm_user.cpp
@@ -395,6 +395,185 @@
return true;
}
+int Worker::ReadUnalignedSector(
+ sector_t sector, size_t size,
+ std::vector<std::pair<sector_t, const CowOperation*>>::iterator& it) {
+ size_t skip_sector_size = 0;
+
+ SNAP_LOG(DEBUG) << "ReadUnalignedSector: sector " << sector << " size: " << size
+ << " Aligned sector: " << it->first;
+
+ if (!ProcessCowOp(it->second)) {
+ SNAP_LOG(ERROR) << "ReadUnalignedSector: " << sector << " failed of size: " << size
+ << " Aligned sector: " << it->first;
+ return -1;
+ }
+
+ int num_sectors_skip = sector - it->first;
+
+ if (num_sectors_skip > 0) {
+ skip_sector_size = num_sectors_skip << SECTOR_SHIFT;
+ char* buffer = reinterpret_cast<char*>(bufsink_.GetBufPtr());
+ struct dm_user_message* msg = (struct dm_user_message*)(&(buffer[0]));
+
+ if (skip_sector_size == BLOCK_SZ) {
+ SNAP_LOG(ERROR) << "Invalid un-aligned IO request at sector: " << sector
+ << " Base-sector: " << it->first;
+ return -1;
+ }
+
+ memmove(msg->payload.buf, (char*)msg->payload.buf + skip_sector_size,
+ (BLOCK_SZ - skip_sector_size));
+ }
+
+ bufsink_.ResetBufferOffset();
+ return std::min(size, (BLOCK_SZ - skip_sector_size));
+}
+
+bool Worker::ReadUnalignedSector(sector_t sector, size_t size) {
+ struct dm_user_header* header = bufsink_.GetHeaderPtr();
+ header->type = DM_USER_RESP_SUCCESS;
+ bufsink_.ResetBufferOffset();
+ std::vector<std::pair<sector_t, const CowOperation*>>& chunk_vec = snapuserd_->GetChunkVec();
+
+ auto it = std::lower_bound(chunk_vec.begin(), chunk_vec.end(), std::make_pair(sector, nullptr),
+ SnapshotHandler::compare);
+
+ // |-------|-------|-------|
+ // 0 1 2 3
+ //
+ // Block 0 - op 1
+ // Block 1 - op 2
+ // Block 2 - op 3
+ //
+ // chunk_vec will have block 0, 1, 2 which maps to relavant COW ops.
+ //
+ // Each block is 4k bytes. Thus, the last block will span 8 sectors
+ // ranging till block 3 (However, block 3 won't be in chunk_vec as
+ // it doesn't have any mapping to COW ops. Now, if we get an I/O request for a sector
+ // spanning between block 2 and block 3, we need to step back
+ // and get hold of the last element.
+ //
+ // Additionally, we need to make sure that the requested sector is
+ // indeed within the range of the final sector. It is perfectly valid
+ // to get an I/O request for block 3 and beyond which are not mapped
+ // to any COW ops. In that case, we just need to read from the base
+ // device.
+ bool merge_complete = false;
+ bool header_response = true;
+ if (it == chunk_vec.end()) {
+ if (chunk_vec.size() > 0) {
+ // I/O request beyond the last mapped sector
+ it = std::prev(chunk_vec.end());
+ } else {
+ // This can happen when a partition merge is complete but snapshot
+ // state in /metadata is not yet deleted; during this window if the
+ // device is rebooted, subsequent attempt will mount the snapshot.
+ // However, since the merge was completed we wouldn't have any
+ // mapping to COW ops thus chunk_vec will be empty. In that case,
+ // mark this as merge_complete and route the I/O to the base device.
+ merge_complete = true;
+ }
+ } else if (it->first != sector) {
+ if (it != chunk_vec.begin()) {
+ --it;
+ }
+ } else {
+ return ReadAlignedSector(sector, size, header_response);
+ }
+
+ loff_t requested_offset = sector << SECTOR_SHIFT;
+
+ loff_t final_offset = 0;
+ if (!merge_complete) {
+ final_offset = it->first << SECTOR_SHIFT;
+ }
+
+ // Since a COW op span 4k block size, we need to make sure that the requested
+ // offset is within the 4k region. Consider the following case:
+ //
+ // |-------|-------|-------|
+ // 0 1 2 3
+ //
+ // Block 0 - op 1
+ // Block 1 - op 2
+ //
+ // We have an I/O request for a sector between block 2 and block 3. However,
+ // we have mapping to COW ops only for block 0 and block 1. Thus, the
+ // requested offset in this case is beyond the last mapped COW op size (which
+ // is block 1 in this case).
+
+ size_t total_bytes_read = 0;
+ size_t remaining_size = size;
+ int ret = 0;
+ if (!merge_complete && (requested_offset >= final_offset) &&
+ (requested_offset - final_offset) < BLOCK_SZ) {
+ // Read the partial un-aligned data
+ ret = ReadUnalignedSector(sector, remaining_size, it);
+ if (ret < 0) {
+ SNAP_LOG(ERROR) << "ReadUnalignedSector failed for sector: " << sector
+ << " size: " << size << " it->sector: " << it->first;
+ return RespondIOError(header_response);
+ }
+
+ remaining_size -= ret;
+ total_bytes_read += ret;
+ sector += (ret >> SECTOR_SHIFT);
+
+ // Send the data back
+ if (!WriteDmUserPayload(total_bytes_read, header_response)) {
+ return false;
+ }
+
+ header_response = false;
+ // If we still have pending data to be processed, this will be aligned I/O
+ if (remaining_size) {
+ return ReadAlignedSector(sector, remaining_size, header_response);
+ }
+ } else {
+ // This is all about handling I/O request to be routed to base device
+ // as the I/O is not mapped to any of the COW ops.
+ loff_t aligned_offset = requested_offset;
+ // Align to nearest 4k
+ aligned_offset += BLOCK_SZ - 1;
+ aligned_offset &= ~(BLOCK_SZ - 1);
+ // Find the diff of the aligned offset
+ size_t diff_size = aligned_offset - requested_offset;
+ CHECK(diff_size <= BLOCK_SZ);
+ if (remaining_size < diff_size) {
+ if (!ReadDataFromBaseDevice(sector, remaining_size)) {
+ return RespondIOError(header_response);
+ }
+ total_bytes_read += remaining_size;
+
+ if (!WriteDmUserPayload(total_bytes_read, header_response)) {
+ return false;
+ }
+ } else {
+ if (!ReadDataFromBaseDevice(sector, diff_size)) {
+ return RespondIOError(header_response);
+ }
+
+ total_bytes_read += diff_size;
+
+ if (!WriteDmUserPayload(total_bytes_read, header_response)) {
+ return false;
+ }
+
+ remaining_size -= diff_size;
+ size_t num_sectors_read = (diff_size >> SECTOR_SHIFT);
+ sector += num_sectors_read;
+ CHECK(IsBlockAligned(sector << SECTOR_SHIFT));
+ header_response = false;
+
+ // If we still have pending data to be processed, this will be aligned I/O
+ return ReadAlignedSector(sector, remaining_size, header_response);
+ }
+ }
+
+ return true;
+}
+
bool Worker::RespondIOError(bool header_response) {
struct dm_user_header* header = bufsink_.GetHeaderPtr();
header->type = DM_USER_RESP_ERROR;
@@ -423,8 +602,7 @@
// Unaligned I/O request
if (!IsBlockAligned(header->sector << SECTOR_SHIFT)) {
- SNAP_LOG(ERROR) << "I/O request is not 4k aligned.";
- return false;
+ return ReadUnalignedSector(header->sector, header->len);
}
return ReadAlignedSector(header->sector, header->len, true);
