--- /dev/null
+* reception stalls permanently (until admin intervention) if the rx buffer
+ queues become empty because of allocation failures (ex. under memory
+ pressure)
+* commit 7c734359d350 ("qlge: Size RX buffers based on MTU.", v2.6.33-rc1)
+ introduced dead code in the receive routines, which should be rewritten
+ anyways by the admission of the author himself, see the comment above
+ ql_build_rx_skb(). That function is now used exclusively to handle packets
+ that underwent header splitting but it still contains code to handle non
+ split cases.
+* truesize accounting is incorrect (ex: a 9000B frame has skb->truesize 10280
+ while containing two frags of order-1 allocations, ie. >16K)
+* while in that area, using two 8k buffers to store one 9k frame is a poor
+ choice of buffer size.
+* in the "chain of large buffers" case, the driver uses an skb allocated with
+ head room but only puts data in the frags.
+* rename "rx" queues to "completion" queues. Calling tx completion queues "rx
+ queues" is confusing.
+* struct rx_ring is used for rx and tx completions, with some members relevant
+ to one case only
+* there is an inordinate amount of disparate debugging code, most of which is
+ of questionable value. In particular, qlge_dbg.c has hundreds of lines of
+ code bitrotting away in ifdef land (doesn't compile since commit
+ 18c49b91777c ("qlge: do vlan cleanup", v3.1-rc1), 8 years ago).
+* triggering an ethtool regdump will hexdump a 176k struct to dmesg depending
+ on some module parameters.
+* the flow control implementation in firmware is buggy (sends a flood of pause
+ frames, resets the link, device and driver buffer queues become
+ desynchronized), disable it by default
+* some structures are initialized redundantly (ex. memset 0 after
+ alloc_etherdev())
+* the driver has a habit of using runtime checks where compile time checks are
+ possible (ex. ql_free_rx_buffers(), ql_alloc_rx_buffers())
+* reorder struct members to avoid holes if it doesn't impact performance
+* in terms of namespace, the driver uses either qlge_, ql_ (used by
+ other qlogic drivers, with clashes, ex: ql_sem_spinlock) or nothing (with
+ clashes, ex: struct ob_mac_iocb_req). Rename everything to use the "qlge_"
+ prefix.
+* avoid legacy/deprecated apis (ex. replace pci_dma_*, replace pci_enable_msi,
+ use pci_iomap)
+* some "while" loops could be rewritten with simple "for", ex.
+ ql_wait_reg_rdy(), ql_start_rx_ring())
+* remove duplicate and useless comments
+* fix weird line wrapping (all over, ex. the ql_set_routing_reg() calls in
+ qlge_set_multicast_list()).
+* fix weird indentation (all over, ex. the for loops in qlge_get_stats())
+* fix checkpatch issues
projects / linux.git / commitdiff