Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

File is getting corrupted when Chunk Mode file transfer interrupted #315

Open
wants to merge 4 commits into
base: master
Choose a base branch
from

Conversation

anamikagsingh
Copy link

Hi,

This change will fix the Chunk mode issue raised in #308

Regards,
Anamika

@norrisjeremy
Copy link
Contributor

Hi @anamikagsingh,

  1. Can you execute mvn formatter:format so that the code changes are formatted correctly?
  2. Can you provide more details on the issue you are seeing and how your proposed changes help to resolve it?
    -- Specifically, can you describe the overall interaction that a JSch consumer would utilize in order to make use of the new checkChunkFailed() & setChunkStart() methods?
    -- It may help if you can provide some short code examples that depict how the file corruption can occur during an interrupted transfer, and then how a user would mitigate the file corruption using the proposed changes.

Thanks,
Jeremy

@anamikagsingh
Copy link
Author

Hi Jeremy,

Could you please let me know if I need to format the file and commit once again or it can be done on committed file too?

In our application, we have an option to transfer large files in smaller chunks and append them in the target in the sequence they have been transferred. But in case the transfer of chunk to the target directory is interrupted due to any issues (i.e., network failure, engine restart etc.), it will transfer the complete chunk again and as a result, the target file gets corrupted.

To solve this, chunk transfer should be resumed instead of transferring the complete chunk again. And, to achieve this, we added a method checkChunkFailed() in ChannelSftp class which checks the bytes which is already transferred and accordingly set the information whether the chunk transfer is interrupted or not. Also, in SFTP application, we have used the below method which will try to resume the chunk transfer if transfer of chunk to the target directory is interrupted.

public void checkChunkTransferInterrupted(byte[] file, String absoluteFilePath, String chunk_start){
String SIGNATURE = "checkChunkTransferInterrupted(byte[] file, String absoluteFilePath, String chunk_start)";
TRACE.entering(SIGNATURE);
InputStream is = new ByteArrayInputStream(file);
try {
	if(sftpChannel.checkChunkFailed(is, absoluteFilePath, chunk_start)){
	setChunkFailed(true, chunk_start);
	TRACE.warningT(SIGNATURE, "Chunk Mode: Chunk transfer was interrupted");
	TRACE.debugT(SIGNATURE, "Chunk Start: " + chunk_start);
	}
} catch (SftpException e) {
	TRACE.warningT(SIGNATURE, "Chunk Mode: Error while checking if chunk transfer interrupted -"+ e.toString()+ ".Continue without check for chunk transfer interruption.");
}
TRACE.exiting(SIGNATURE);
}

public void setChunkFailed(boolean chunk_failed, String chunk_start){
this.chunk_failed = chunk_failed;
sftpChannel.setChunkStart(chunk_start);
}

@anamikagsingh
Copy link
Author

Hi Jeremy,

I also have added changed files 'Session.java' and 'UserAuthKeyboardInteractive.java' to this pull request. This change is to fix the issue raised in #302 and also to support multiple server fingerprint.

Regards,
Anamika

@norrisjeremy
Copy link
Contributor

Hi @anamikagsingh,

  1. You need to execute mvn formatter:format and commit, then push the resulting changes to this PR.
  2. Can you please split the changes to Session.java and UserAuthKeyboardInteractive.java into individual commits, instead of a single combined commit?
  3. Can you modify your commit messages to be more descriptive than simply ChannelSftp and SessionAndUserAuthKeyboardInteractive?
  4. I'm not sure I follow what you mean by we have an option to transfer large files in smaller chunks and append them in the target in the sequence they have been transferred: how are you transferring these large files via smaller chunks using JSch's existing ChannelSftp API methods?
    -- I also do not understand how an application would determine the chunkStart value to use for the checkChunkFailed() method: can you better explain how you are determining this value?
  5. I'm not sure I understand the purpose of the changes to Session.java: can't you implement this custom logic of authenticating hosts by their host key fingerprint by simply implementing your own custom HostKeyRepository and then instructing JSch use it (via either jsch.setHostKeyRepository() or session.setHostKeyRepository())?
  6. I'm not sure I understand the proposed changes to UserAuthKeyboardInteractive.java: why not simply implement a custom UserInfo as suggested in UnsupportedOperationException occurs when password prompt is in a different format #302?
    -- Also, you mentioned in UnsupportedOperationException occurs when password prompt is in a different format #302 that you had a server that presented a prompt of Enter password:, but doesn't that prompt already work without any changes since that should already match the existing prompt[0].toLowerCase().indexOf("password:") >= 0?

Thanks,
Jeremy

src/main/java/com/jcraft/jsch/Session.java Outdated Show resolved Hide resolved
src/main/java/com/jcraft/jsch/Session.java Outdated Show resolved Hide resolved
src/main/java/com/jcraft/jsch/Session.java Outdated Show resolved Hide resolved
@@ -129,7 +129,8 @@ public boolean start(Session session) throws Exception {
byte[][] response = null;

if (password != null && prompt.length == 1 && !echo[0]
&& prompt[0].toLowerCase().indexOf("password:") >= 0) {
&& (prompt[0].toLowerCase().indexOf("password:") >= 0
|| prompt[0].toLowerCase().indexOf("password") >= 0)){
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In #302 you mention that this change is needed in order to support a server that has a prompt of Enter password:.
However, the string Enter password: already should match prompt[0].toLowerCase().indexOf("password:") >= 0, so I'm not sure I understand the purpose of this change (see below)?

jshell> String foo = "Enter password:";
foo ==> "Enter password:"

jshell> if (foo.toLowerCase().indexOf("password:") >= 0) System.out.println("matches"); else throw new Exception("fail");
matches

jshell> 

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I am not able to reproduce this issue in my test system so cannot test it. We can leave the changes done in UserAuthKeyboardInteractive.java for now. Please let me know if I need to revert the changes and commit again. Thanks.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If you are not able to reproduce the issue that you believed was fixed by this change, then it should be removed.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The or also overlaps, you could just remove the colon..

@@ -643,6 +653,32 @@ public void _put(InputStream src, String dst, SftpProgressMonitor monitor, int m
throw (SftpException) e;
throw new SftpException(SSH_FX_FAILURE, e.toString(), e);
}

public boolean checkChunkFailed(InputStream src, String dst, String ChunkStart)throws SftpException{
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. The argument InputStream src appears to be unused in this method, so why is it needed?
  2. The argument String ChunkStart should be changed to not start with a capital letter.
    -- Perhaps something like String chunkStart instead?

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. Removed the argument InputStream src from the method.
  2. Changed the parameter name to startChunk

@@ -32,6 +32,9 @@
import java.util.Hashtable;
import java.util.Vector;

import com.jcraft.jsch.Channel.MyPipedInputStream;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is unnecessary.
Since ChannelSftp extends Channel, it can directly reference the MyPipedInputStream class without an import statement.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Removed the import as it was not needed. Thanks for the suggestions.

throw new SftpException(SSH_FX_FAILURE, "failed to resume for "+dst);
}
}
else if(mode==RESUME && skip>0){
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this could be better written as simply something like this to avoid code duplication:

      if (mode == RESUME && skip > 0) {
        long srcSkip = chunkStart != null ? skip - Long.parseLong(chunkStart) : skip;
        long skipped = src.skip(srcSkip);
        if (skipped < srcSkip) {
          throw new SftpException(SSH_FX_FAILURE, "failed to resume for " + dst);
        }
      }

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have made the changes as suggested.

@@ -158,6 +161,7 @@ public class ChannelSftp extends ChannelSession {
private boolean fEncoding_is_utf8 = true;

private RequestQueue rq = new RequestQueue(16);
private String chunkStart = null;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why does chunkStart need to be a String instead of having it simply be a long?
This would avoid a lot of unwieldy calls to Long.parseLong(chunkStart)?

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Changed it to Long now and made the other changes accordingly.

break;
}
}
}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Also, can you point at any other applications or libraries that perform SSH host key verification using fingerprints?
This seems like a pretty unusual (and potentially insecure) practice that I am unaware of, and I'm not sure we want to directly enable this sort of thing natively within JSch.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have no idea if any other application which is using Jsch library is also using fingerprints to perform SSH host key verification. It is being used in SAP PI SFTP Adapter.

@@ -129,6 +129,8 @@ public class Session {

Buffer buf;
Packet packet;

List<String> fingerprintList = new ArrayList<String>();
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This can be shortened to simply new ArrayList<>() instead of new ArrayList<String>().

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Changed it as suggested.

@norrisjeremy
Copy link
Contributor

HI @anamikagsingh,

Can you please answer the following questions?

  1. I'm not sure I follow what you mean by we have an option to transfer large files in smaller chunks and append them in the target in the sequence they have been transferred: how are you transferring these large files via smaller chunks using JSch's existing ChannelSftp API methods? Can you provide some example code of how you are using ChannelSftp and how your proposed API changes are utilized to remedy the file corruption issue you are having?
  2. I'm not sure I understand the purpose of the changes to Session.java: is there a reason you cannot simply achieve this custom logic of authenticating hosts by their host key fingerprint by simply implementing your own custom HostKeyRepository and then instructing JSch use it (via either jsch.setHostKeyRepository() or session.setHostKeyRepository())?

Thanks,
Jeremy

@anamikagsingh
Copy link
Author

HI @anamikagsingh,

Can you please answer the following questions?

  1. I'm not sure I follow what you mean by we have an option to transfer large files in smaller chunks and append them in the target in the sequence they have been transferred: how are you transferring these large files via smaller chunks using JSch's existing ChannelSftp API methods? Can you provide some example code of how you are using ChannelSftp and how your proposed API changes are utilized to remedy the file corruption issue you are having?
  2. I'm not sure I understand the purpose of the changes to Session.java: is there a reason you cannot simply achieve this custom logic of authenticating hosts by their host key fingerprint by simply implementing your own custom HostKeyRepository and then instructing JSch use it (via either jsch.setHostKeyRepository() or session.setHostKeyRepository())?

Thanks, Jeremy

About point 1, when chunkmode is enabled, we set the chunk size (i.e. 10 MB). Based on chunk size, the bytes will be read from the input file and processed further. We are using the below snippet for chunk mode transfer in Adapter:


  if(chunkMode) {					
	long readBytes = 0;
	String user = propHandler.getPropertyValueAsString(ChannelConstants.AUTH_USERNAME);
	String chunkID = getChunkID(con.getHost(), user, fileAtts);
	//for recovery in chunk mode
	boolean retry = false;
	String chunkEntry = checkChunkStatus(chunkID);
	if(chunkEntry != null) {
		readBytes = Long.parseLong(chunkEntry);
		if(readBytes > 0) {
		retry = true;
		 }
		}				
		//read from inputstream
		byte[] chunk = new byte[chunkSize];
		int curChunkSize = -1;
		curChunkSize = readChunk(chunk, is, chunkSize);
		TRACE.debugT(SIGNATURE, "Bytes read: " + curChunkSize);
		if(curChunkSize >0 || processEmptyFile ==true) {
		audit.addAuditLogEntry(msg.getMessageKey(), AuditLogStatus.SUCCESS, "Starting to read file from byte: " + readBytes);
		readBytes += curChunkSize;
		TRACE.debugT(SIGNATURE, "Total read Bytes: " + readBytes);

		TxTicket ticket = null;
		//sending message to XI
		try {
			//create transaction
			ticket = createTransaction(channelDeliverySemantics);
			updateChunkStatus(chunkID, readBytes + "");
			if(curChunkSize == chunkSize) {
				sendMessageFromSftp(chunk, fileAtts, attachments, channel, con, home, msg,isSpecialFileHandling, specialFileStorageLocation);
			} else {
				byte[] trimChunk = new byte[curChunkSize];
				System.arraycopy(chunk, 0, trimChunk, 0, curChunkSize);
				sendMessageFromSftp(trimChunk, fileAtts, attachments, channel, con, home, msg,isSpecialFileHandling, specialFileStorageLocation);
				
				//final chunk; hence add it to message ID mapper for duplicate check
				addToIDMapper(con.getHost(), fileAtts, msg.getMessageKey().getMessageId());
			}
			if(ticket != null) {
				commitTransaction(ticket);
			}
			xiMsgSent = true;
		} catch(MessagingException e) {
		 }
	}                               

About point 2, I will try to figure it out.

Thanks.

@norrisjeremy
Copy link
Contributor

Hi @anamikagsingh,

From what I can understand, this chunkmode handling seems to be specific to your application.
As such, I'm not sure if I believe it is appropriate to merge it into JSch, lest it introduce regressions for other JSch users.

Thanks,
Jeremy

Copy link

sonarcloud bot commented Mar 15, 2024

Quality Gate Passed Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
No data about Coverage
No data about Duplication

See analysis details on SonarCloud

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants