• course outline
• .o, .a, .so
• building

• make

• kernel mode
• system calls

• exceptions
  • events triggering
  • what the hardware must do
  • exception handlers
• threads and context switches
• process = threads + address space

• Unix signals
  • events triggering
  • signal handlers and implementation

• user IDs/account abstraction
• POSIX permissions
• sudo is not kernel mode

• vulnerabilities and privilege escalation

• virtual memory concept
• page tables
  • concept, small example

• page tables
  • storing in memory
  • in-memory lookup example
  • as a tree (start)

• page tables
  • as a tree (finish)
  • in-memory page table lookup
  • permission bits
  • page table tricks, loading data on demand, swapping

• mailbox analogy
  • inboxes and outboxes
  • addresses
  • best-effort delivery
  • envelope versus contents
• TCP / connections
  • acknowledgements
  • state machine
• names and addresses
  • layers preview

• example: a frame
  • layers (link, network, transport)
• IP
  • addresses (v4, v6)
  • routers
• port numbers and UDP (start)

• port numbers and UDP (finish)
• DNS
  • is a distributed database
  • with caching
• URLs
• (if time) link-layer to network-layer mapping
• DHCP
• NAT

• cryptography (start)
  • symmetric encryption
  • asymmetric encryption
  • digital signature
• replay attacks

• cryptography (con’t)
  • cryptographic hashes
  • key exchange
• public key infrastructure
• TLS protocol outline

• connecting processors to memory
• memory hierarchy
• locality
• direct-mapped cache
  • valid bits
  • tags

• cache misses and C code

• set-associative cache
• replacement policies

(Guest lecture; Prof Reiss will be out of town.)

• review of set-associative caches
• write policies
• three Cs, cache tradeoffs (start)

• cache tradeoffs, very briefly

• the TLB

• processes, threads
• creating new processes
  • fork, exec*
• waiting for processes

• file descriptors
• pipe and dup2

• creating new threads
• thread use examples
• problem of synchronization

• atomicity, race conditions
  • atomic read-modify-write instructions
• locks and mutexes

• barriers

• interlude: reordering

• deadlock examples

• deadlock con’t

• monitors (mutexes + condition variables) (start)

• monitors (finish)
• (briefly) counting semaphores
• (if time) reader/writer locks
• transactions (briefly)

• pipelining idea
• review: single-cycle CPU
• pipelining for the CPU (
• data hazards
• hazard handling, part 1
• stalling / waiting

• hazard handling, part 2
• forwarding and forwarding logic
• control hazards + speculation idea (start)
  • guess and check later

• speculation idea (finish)
  • guess and check later
  • undo when wrong
• branch prediction ideas
  • forward-taken, backwards-not-taken
  • using history

• interlude: buses and direct-memory access

• idea of multiple issue + out-of-order
• out-of-order pipeline
  • aside: predicting branch targets
• register renaming
• instruction queues (start)

• instruction queues (finish)

• data flow model + thinking about OOO performance

• side channels generally
• side channels in the cache (start)

• side channels in the cache (finish)
• meltdown:
  • speculation past a priv/protection fault
• spectre
  • speculation past a bounds chcek
  • using cache side-channels

notes

9am-12pm.

Chemistry 402.