From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 25376 invoked by alias); 1 Oct 2007 05:01:14 -0000 Received: (qmail 25359 invoked by uid 22791); 1 Oct 2007 05:01:12 -0000 X-Spam-Check-By: sourceware.org Received: from pop132.ocn.ne.jp (HELO pop132.ocn.ne.jp) (60.37.31.215) by sourceware.org (qpsmtpd/0.31) with ESMTP; Mon, 01 Oct 2007 05:01:05 +0000 Received: from ariga (p1143-ipbf2202marunouchi.tokyo.ocn.ne.jp [122.17.131.143]) by pop132.ocn.ne.jp (OCN) with SMTP id l91511BT024200; Mon, 1 Oct 2007 14:01:01 +0900 (JST) Message-ID: <000301c803e7$fd728a50$1c0110ac@ariga> From: "ariga masahiro" To: "Andrew Lunn" Cc: References: <001501c801b7$2de8bed0$1c0110ac@ariga> Date: Mon, 01 Oct 2007 05:01:00 -0000 MIME-Version: 1.0 Content-Type: text/plain; format=flowed; charset="iso-8859-1"; reply-type=response Content-Transfer-Encoding: 7bit X-Mailer: Microsoft Outlook Express 6.00.2900.2869 X-IsSubscribed: yes Mailing-List: contact ecos-discuss-help@ecos.sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: ecos-discuss-owner@ecos.sourceware.org Subject: [ECOS] Not working lan91cxx_sc drv X-SW-Source: 2007-10/txt/msg00001.txt.bz2 Hi, I am the same person who posted "Question about interrupt trigger mode",dated 9/28/2007 but since then I am encountered the most serious problem so I change title and post this most-urgent,most-ernest and most-desperate mail. As I said, My target uses SMSC LAN91C111 chip,CPU is SH7709S. As ethernet interrupt,uses IRQ3 line. I build "net" template,eCos source is updated by CVS checkout. As I inserted LAN91C111's mask routine in the ISR, I succeeded to entered into DSR. (This interrupt was caused not by eCos-proper code, but by my tampered routine.I will later explain.) But although RCV-INT is assserted,delivering-packet routine never called, so I entered repetedly ISR and DSR. Then I discovered LAN91C111 driver's receive routine was never called. Please forgive me long mail,but I try to inform you as correctly as possible. First I relate current problem,and later describe what I have tampered with. I register SMSC interrupt ruoitne in \packages\devs\eth\smsc\lan91cxx\current\src\if_lan91cxx.c's smsc_lan91cxx_init(struct cyg_netdevtab_entry *tab) like this, #ifndef CYGPKG_IO_ETH_DRIVERS_STAND_ALONE // Initialize environment, setup interrupt handler cyg_drv_interrupt_create(cpd->interrupt, CYGNUM_DEVS_ETH_SMSC_LAN91CXX_INT_PRIO, (cyg_addrword_t)sc, // Data item passed to interrupt handler (cyg_ISR_t *)lan91cxx_isr, (cyg_DSR_t *)eth_drv_dsr, // The logical driver DSR &lan91cxx_interrupt_handle, &lan91cxx_interrupt); cyg_drv_interrupt_attach(lan91cxx_interrupt_handle); #endif // !CYGPKG_IO_ETH_DRIVERS_STAND_ALONE cyg_drv_interrupt_acknowledge(cpd->interrupt); cyg_drv_interrupt_unmask(cpd->interrupt); And register LAN91C111 driver in my target's \packages\devs\eth\sh\inserter\current\include\devs_eth_inserter.inl. I think this is important so I put on all content. #include #include #include // MAC address is stored as a Redboot config option #ifdef CYGPKG_REDBOOT #include #ifdef CYGSEM_REDBOOT_FLASH_CONFIG #include #include #define LAN91CXX_IS_LAN91C111 RedBoot_config_option("Network hardware address [MAC]", inserter_esa, ALWAYS_ENABLED, true, CONFIG_ESA, 0 ); #endif #endif // ESA address fetch function static void inserter_get_ESA(struct lan91cxx_priv_data *cpd) { // Fetch hardware address from RedBoot config #if defined(CYGSEM_DEVS_ETH_SH_INSERTER_REDBOOT_ESA) #if defined(CYGPKG_REDBOOT) && \ defined(CYGSEM_REDBOOT_FLASH_CONFIG) flash_get_config("inserter_esa", cpd->enaddr, CONFIG_ESA); #else #error "No RedBoot flash configuration to store ESA" #endif #else unsigned char static_esa[] = CYGDAT_DEVS_ETH_SH_INSERTER_ESA; memcpy(cpd->enaddr, static_esa, 6); #endif } static lan91cxx_priv_data lan91cxx_eth0_priv_data = { config_enaddr : inserter_get_ESA, #ifndef CYGSEM_DEVS_ETH_SH_INSERTER_REDBOOT_ESA enaddr: CYGDAT_DEVS_ETH_SH_INSERTER_ESA, hardwired_esa : true, #else hardwired_esa : false, #endif #if 0 base : (unsigned short *) SA1110_FHH_ETH_IOBASE, attbase : (unsigned char *) SA1110_FHH_ETH_MMBASE, interrupt : SA1110_IRQ_GPIO_ETH #else base : (unsigned short *) 0xa8000000, interrupt : 9, #endif }; ETH_DRV_SC(lan91cxx_sc, &lan91cxx_eth0_priv_data, // Driver specific data CYGDAT_DEVS_ETH_SH_INSERTER_NAME, // Name for device lan91cxx_start, lan91cxx_stop, lan91cxx_control, lan91cxx_can_send, lan91cxx_send, lan91cxx_recv, lan91cxx_deliver, lan91cxx_poll, lan91cxx_int_vector ); NETDEVTAB_ENTRY(lan91cxx_netdev, "lan91cxx_" CYGDAT_DEVS_ETH_SH_INSERTER_NAME, smsc_lan91cxx_init, &lan91cxx_sc); //EOF devs_eth_inserter.inl I found ETH_DRV_SC definition as below. \packages\io\eth\current\include\eth_drv.h #define ETH_DRV_SC(sc,priv,name,start,stop,control,can_send,send,recv,deliver,poll,int_vector) \ static void start(struct eth_drv_sc *sc, unsigned char *enaddr, int flags); \ static void stop(struct eth_drv_sc *sc); \ static int control(struct eth_drv_sc *sc, unsigned long key, void *data, int data_length); \ static int can_send(struct eth_drv_sc *sc); \ static void send(struct eth_drv_sc *sc, struct eth_drv_sg *sg_list, int sg_len, int total, unsigned long key); \ static void recv(struct eth_drv_sc *sc, struct eth_drv_sg *sg_list, int sg_len); \ static void deliver(struct eth_drv_sc *sc); \ static void poll(struct eth_drv_sc *sc); \ static int int_vector(struct eth_drv_sc *sc); \ static struct eth_hwr_funs sc##_funs = { \ start, \ stop, \ control, \ can_send, \ send, \ recv, \ deliver, \ poll, \ int_vector, \ ð_drv_funs, \ (struct eth_drv_funs *)0 }; \ struct eth_drv_sc sc = {&sc##_funs, priv, name}; I assume this is developed like below, #define ETH_DRV_SC(lan91cxx_sc,&lan91cxx_eth0_priv_data,CYGDAT_DEVS_ETH_SH_INSERTER_NAME,lan91cxx_start,lan91cxx_stop,lan91cxx_control,lan91cxx_can_send,lan91cxx_send,lan91cxx_recv,lan91cxx_deliver,lan91cxx_poll,lan91cxx_int_vector) \ static void lan91cxx_start(struct eth_drv_sc *lan91cxx_sc, unsigned char *enaddr, int flags); \ static void lan91cxx_stop(struct eth_drv_sc *lan91cxx_sc); \ static int lan91cxx_control(struct eth_drv_sc *lan91cxx_sc, unsigned long key, void *data, int data_length); \ static int lan91cxx_can_send(struct eth_drv_sc *lan91cxx_sc); \ static void lan91cxx_send(struct eth_drv_sc *lan91cxx_sc, struct eth_drv_sg *sg_list, int sg_len, int total, unsigned long key); \ static void lan91cxx_recv(struct eth_drv_sc *lan91cxx_sc, struct eth_drv_sg *sg_list, int sg_len); \ static void lan91cxx_deliver(struct eth_drv_sc *lan91cxx_sc); \ static void lan91cxx_pollpoll(struct eth_drv_sc *lan91cxx_sc); \ static int lan91cxx_int_vector(struct eth_drv_sc *lan91cxx_sc); \ static struct eth_hwr_funs lan91cxx_sc_funs = { \ lan91cxx_start, \ lan91cxx_stop, \ lan91cxx_control, \ lan91cxx_can_send, \ lan91cxx_send, \ lan91cxx_recv, \ lan91cxx_deliver, \ lan91cxx_poll, \ lan91cxx_int_vector, \ ð_drv_funs, \ (struct eth_drv_funs *)0 }; \ struct eth_drv_sc lan91cxx_sc = {&lan91cxx_sc_funs, &lan91cxx_eth0_priv_data, CYGDAT_DEVS_ETH_SH_INSERTER_NAME}; I assume lan91cxx_recv must be called to operate on packets, but it was never called. I traced DSR routine DSR is this \packages\io\eth\current\src\net\eth_drv.c void eth_drv_dsr(cyg_vector_t vector, cyg_ucount32 count, cyg_addrword_t data) { struct eth_drv_sc *sc = (struct eth_drv_sc *)data; #ifdef CYGDBG_USE_ASSERTS // then check that this really is a "sc" { cyg_netdevtab_entry_t *t; for (t = &__NETDEVTAB__[0]; t != &__NETDEVTAB_END__; t++) if ( ((struct eth_drv_sc *)t->device_instance) == sc ) break; // found it CYG_ASSERT( t != &__NETDEVTAB_END__, "eth_drv_dsr: Failed to find sc in NETDEVTAB" ); } #endif // Checking code sc->state |= ETH_DRV_NEEDS_DELIVERY; ecos_synch_eth_drv_dsr(); // [request] run delivery function for this dev } And in \packages\net\bsd_tcpip\current\src\ecos\timeout.c void ecos_synch_eth_drv_dsr(void) { cyg_flag_setbits( &alarm_flag, 2 ); } And in \packages\kernel\current\src\common\kapi.cxx void cyg_flag_setbits( cyg_flag_t *flag, cyg_flag_value_t value) __THROW { ((Cyg_Flag *)flag)->setbits( value ); } And in \packages\kernel\current\src\sync\flag.cxx void Cyg_Flag::setbits( Cyg_FlagValue arg ) { CYG_REPORT_FUNCTION(); CYG_ASSERTCLASS( this, "Bad this pointer"); // Prevent preemption Cyg_Scheduler::lock(); // OR in the argument to get a new flag value. value |= arg; // anyone waiting? if ( !(queue.empty()) ) { FlagWaitInfo *p; Cyg_Thread *thread; Cyg_ThreadQueue holding; do { thread = queue.dequeue(); p = (FlagWaitInfo *)(thread->get_wait_info()); CYG_ASSERT( (p->allmask == 0) != (p->anymask == 0), "Both masks set" ); CYG_ASSERT( 0 == p->value_out, "Thread already awoken?" ); if ( ((p->allmask != 0) && (p->allmask & value) == p->allmask) || ((p->anymask & value) != 0 ) ) { // success! awaken the thread thread->set_wake_reason( Cyg_Thread::DONE ); thread->wake(); // return the successful value to it p->value_out = value; // do we clear the value; is this the end? if ( p->do_clear ) { // we can break here but need to preserve ordering value = 0; // so let it cycle the whole queue regardless } } else { // preserve the entry on the holding queue holding.enqueue( thread ); } } while ( !(queue.empty()) ); // Now re-queue the unaffected threads back into the flag queue while ( !(holding.empty()) ) { queue.enqueue( holding.dequeue() ); } } // Unlock scheduler and allow other threads to run Cyg_Scheduler::unlock(); CYG_REPORT_RETURN(); } Truely,I cannot fathom this coding and I would appreciate if you kindly teach me how lan91cxx_recv is to be called. Now I describe what I have tampered with in detail,please forgive me lengthy details. The beginning of the matter is when I ran eCos as it is,I couldn't make LAN91C111 into LINK State, LINK LED didn't light.So I inserted my own routine in order to operate LAN91C111 into LINK State in top of \packages\io\eth\current\src\net\eth_drv.c's eth_drv_init function like below. eth_drv_init(struct eth_drv_sc *sc, unsigned char *enaddr) { // this is my tampered coding cyg_netdevtab_entry_t *t; struct ifnet *ifp = &sc->sc_arpcom.ac_if; #ifdef CYGPKG_NET_FREEBSD_STACK int unit; char *np, *xp; #endif // this is my tampered coding for (t = &__NETDEVTAB__[0]; t != &__NETDEVTAB_END__; t++) { log(LOG_INIT, "Init device '%s'\n", t->name); if (smsc_91c111_init(t)) { // this is my concocted routine t->status = CYG_NETDEVTAB_STATUS_AVAIL; } else { // What to do if device init fails? t->status = 0; // Device not [currently] available } } and in order making LAN91C111 into LINK State it was necessary to use interrupt so I enabled LAN91C111 interrupt in my concocted routine. Above-mentioned interrupt was caused by my tampered code.At LINK-established time interrupt happened. On second thought, I realized my tampering was wrong. So I quitted my tampering routine, I returned to eCos original source and checked why couldn't make LINK LED lighed. I discovered it never called lan91cxx routines(i.e. ETH_DRV_SC routines). Below is RedBoot output messages log. I breaked lan91cxx_start() but never entered it. I made DEBUG_FUNCTION() available. -- RedBoot output messages log My Flash ID is 4:22f9:0:19 eth_drv_init:enaddr=0x8c0005c4 Ethernet eth0: MAC address 00:40:31:08:01:00 IP: 172.16.1.200/255.255.255.0, Gateway: 172.16.1.1 Default server: 172.16.1.1 RedBoot(tm) bootstrap and debug environment [ROM] Non-certified release, version UNKNOWN - built 16:50:29, Sep 19 2007 Platform: inserter (SH 7709S) Copyright (C) 2000, 2001, 2002, 2003, 2004 Red Hat, Inc. Copyright (C) 2003, 2004, 2005, 2006 eCosCentric Limited RAM: 0x8c000000-0x90000000, [0x8c00bed0-0x8ffed000] available FLASH: 0xa0000000 - 0xa0400000, 64 blocks of 0x00010000 bytes each. == Executing boot script in 3.000 seconds - enter ^C to abort RedBoot> fis load -b 0x8c010000 nc_slave RedBoot> channel 1 RedBoot> go 0x8c010000 Network stack using 69632 bytes for misc space 69632 bytes for mbufs 139264 bytes for mbuf clusters [cyg_net_init] Init: mbinit(0x00000000) [cyg_net_init] Init: cyg_net_init_devs(0x00000000) Init device 'lan91cxx_eth0' smsc_lan91cxx_init LAN91CXX - supposed BankReg @ a800000e = 3302 LAN91CXX - type: 9, rev: 1 LAN91CXX - status: 0069 LAN91CXX - static ESA: 00:40:31:08:01:00 [cyg_net_init] Init: loopattach(0x00000000) [cyg_net_init] Init: ifinit(0x00000000) [cyg_net_init] Init: domaininit(0x00000000) [cyg_net_init] Init: cyg_net_add_domain(0x8c05a7b4) New domain internet at 0x00000000 [cyg_net_init] Init: cyg_net_add_domain(0x8c05a1f8) New domain route at 0x00000000 [cyg_net_init] Init: call_route_init(0x00000000) [cyg_net_init] Done Start Network Characterization - SLAVE No load = 58470 Set background load = 20% Set background load = 0% High Load[20] = 37131 => 37% Set background load = 20% Set background load = 0% Load[10] = 47736 => 19% Set background load = 20% Set background load = 0% Final load[10] = 47853 => 19% Start test for eth0 -- end of RedBoot output messages log All matters considered,I deduce that ETH_DRV_SC-table-reference-functions are not working. I cannot understand where and how they are called. Please teach me how to correctly call these functions. Particularly I am curious to know eCos's system contrivance for calling drv routines and if you know anything(e.g. web site,etc.) to guide me, please enlighten me. I am "a drowning man catching at a straw". I admit my knowledge is below than ecos level, but my company's expecting in applying eCos into products. And I trust eCos, so please do not let me down and I eanestly beseech your kind advice. Masahiro Ariga -- Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss