(Good grief. My earlier, equally lengthy response didn't seem to post. I think the information is important enough to resolving this to reconstruct it...)
The XBD files are the latest from the Avnet website: avnet_edk13_3_xbd_files_11_11_2011.zip.
The UCF in question is (22 June, 2010)
avnet_edk13_3_xbd_files/Avnet/boards/Avnet_V5FXT_Evaluation_board/data/Hard_Ethernet_MAC_xps_ll_temac.ucf.
line 30:
------------
# EMAC0 RX PHY Clock
NET "*/GMII_RX_CLK_0*" TNM_NET = "phy_clk_rx0";
I'm a bit befuddled to find only one of the three NETs defined here, and it isn't an exact match. From where else does XPS gather constraints in system.ucf?
The signals appear to be present in the pcores.
In $XILINX_EDK/hw/XilinxProcessorIPLib/pcores/xps_ll_temac_v2_03_a, searching for the signal names finds reasonable looking matches.
$ grep -i "rx_client_clk" `find . -type f`
...
./hdl/vhdl/v5_single_gmii_top.vhd: RX_CLIENT_CLK_ENABLE_0 : out std_logic;
./hdl/vhdl/v5_single_gmii_top.vhd: signal rx_client_clk_out_0_i : std_logic;
./hdl/vhdl/v5_single_gmii_top.vhd: signal rx_client_clk_in_0_i : std_logic;
./hdl/vhdl/v5_single_gmii_top.vhd: rx_client_clk_in_0_i <= gmii_rx_clk_0_i;
...
./hdl/vhdl/xps_ll_temac.vhd:signal rX_CLIENT_CLK_0 : std_logic;
./hdl/vhdl/xps_ll_temac.vhd:signal rX_CLIENT_CLK_1 : std_logic;
./hdl/vhdl/xps_ll_temac.vhd: Temac0AvbRxClk <= rX_CLIENT_CLK_0;
./hdl/vhdl/xps_ll_temac.vhd:RxClientClk_0 <= rX_CLIENT_CLK_0;
./hdl/vhdl/xps_ll_temac.vhd:RxClientClk_1 <= rX_CLIENT_CLK_1;
./hdl/vhdl/xps_ll_temac.vhd: rxClClk => rX_CLIENT_CLK_0,--: in
...
$ grep -i "tx_gmii_mii_clk" `find . -type f`
...
./hdl/vhdl/v5_single_gmii_top.vhd: signal tx_gmii_mii_clk_out_0_i : std_logic;
./hdl/vhdl/v5_single_gmii_top.vhd: signal tx_gmii_mii_clk_in_0_i : std_logic;
./hdl/vhdl/v5_single_gmii_top.vhd: TX_CLK => tx_gmii_mii_clk_in_0_i,
./hdl/vhdl/v5_single_gmii_top.vhd: tx_gmii_mii_clk_in_0_i <= TX_CLK_0;
...
./hdl/vhdl/v5_temac_wrap.vhd: signal tx_gmii_mii_clk0_temp : std_logic;
./hdl/vhdl/v5_temac_wrap.vhd: signal tx_gmii_mii_clk1_temp : std_logic;
...
$ grep -i "GMII_RX_CLK_0" `find . -type f`
...
./hdl/vhdl/v5_single_gmii_top.vhd: GMII_RX_CLK_0 : in std_logic;
./hdl/vhdl/v5_single_gmii_top.vhd: signal gmii_rx_clk_0_i : std_logic;
./hdl/vhdl/v5_single_gmii_top.vhd: RX_CLK => gmii_rx_clk_0_i);
./hdl/vhdl/v5_single_gmii_top.vhd: gmii_rx_clk_0_i <= GMII_RX_CLK_0;
...
./hdl/vhdl/xps_ll_temac.vhd: GMII_RX_CLK_0 => GMII_RX_CLK_0, -- in
./hdl/vhdl/xps_ll_temac.vhd: GMII_RX_CLK_0 => GMII_RX_CLK_0, -- in
./hdl/vhdl/xps_ll_temac.vhd: GMII_RX_CLK_0 => GMII_RX_CLK_0, -- in
./hdl/vhdl/xps_ll_temac.vhd: GMII_RX_CLK_0 => GMII_RX_CLK_0, -- in
As though that did anything to hide the old definitions. Deleting the old XBD tree cleared the problem.