Changes from 1.1.51 to 1.1.52

These changes are archived on ftp.crynwr.com:pub/kchanges and http://www.crynwr.com/kchanges. I cannot answer questions about Linux -- I merely summarize the kernel patches after reading them. I try to make them useful to everyone, but kernel hackers should take them with a grain of salt and read the patches themselves. The patch file is patch52.gz

• Updated kernel-contributors list.
• Big changes in floppy controller (too many for me to understand).
• IDE driver changes for non-disk devices.
• Ramdisk driver can now ask the user for the ramdisk.
• Console driver now uses memory at 0xa0000 for black&white maps.
• lp driver now modularized.
• Change to line discipline -- if EOF char, set TTY_PUSH in flags (gets rid of gobble_eof).
• tty driver now checks for read permission on ioctls.
• PLIP driver uses defines instead of literals.
• If the filesystem had to write a buffer, search all the lists again.
• If the filename being executed is invalid, be sure to return an error.
• Buffering added to ext2 fs.
• Locking was checking for write permission on its arguments, not read.
• NFS changes:
  • Do not copy data too often around in the kernel.
  • In nfs_file_read the return value of kmalloc wasn't checked.
  • Put in a better version of read look-ahead buffering. Original idea and implementation by Wai S Kok elekokw@ee.nus.sg.
  • incorrect XID used to confuse Linux.
  • use the correct max size, if reading a packet and also verify, if the whole packet has been read...
  • more checks should be done in proc.c...
• Superblock handler now uses a bitmap for tracking devices.
• Improvements to sysv/coherent filesystem (now looks for sysv4).
• Permit PCI BIOSes with two Service Directories.
• System call for interval timers didn't check for read permission on
• its memory.
• Add a capability to limit a process's cpu time.
• time code in kernel wasn't setting the microseconds field to zero everywhere.
• Kernel wasn't checking for read permission in stime or settimeofday syscalls.
• Skip past '*' character in kernel's vsprintf code.
• Read-only virtual memory wasn't implemented quite right.
• Some icmp packets were not being put on right device's transmit queue.
• Update the loopback route only if there was already one.

[Index][Next]