X-Git-Url: https://err.no/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=drivers%2Fnet%2Fcassini.c;h=f1936d51b458e387e3ba0ffef867f1e388bb8a90;hb=057b61afca098d3ad3d9e8d15914bc9f9315e425;hp=7df31b5561cc02000cbc44d1ea2048bd016f4e58;hpb=0b776eb5426752d4e53354ac89e3710d857e09a7;p=linux-2.6 diff --git a/drivers/net/cassini.c b/drivers/net/cassini.c index 7df31b5561..f1936d51b4 100644 --- a/drivers/net/cassini.c +++ b/drivers/net/cassini.c @@ -142,8 +142,8 @@ #define DRV_MODULE_NAME "cassini" #define PFX DRV_MODULE_NAME ": " -#define DRV_MODULE_VERSION "1.4" -#define DRV_MODULE_RELDATE "1 July 2004" +#define DRV_MODULE_VERSION "1.6" +#define DRV_MODULE_RELDATE "21 May 2008" #define CAS_DEF_MSG_ENABLE \ (NETIF_MSG_DRV | \ @@ -336,30 +336,6 @@ static inline void cas_mask_intr(struct cas *cp) cas_disable_irq(cp, i); } -static inline void cas_buffer_init(cas_page_t *cp) -{ - struct page *page = cp->buffer; - atomic_set((atomic_t *)&page->lru.next, 1); -} - -static inline int cas_buffer_count(cas_page_t *cp) -{ - struct page *page = cp->buffer; - return atomic_read((atomic_t *)&page->lru.next); -} - -static inline void cas_buffer_inc(cas_page_t *cp) -{ - struct page *page = cp->buffer; - atomic_inc((atomic_t *)&page->lru.next); -} - -static inline void cas_buffer_dec(cas_page_t *cp) -{ - struct page *page = cp->buffer; - atomic_dec((atomic_t *)&page->lru.next); -} - static void cas_enable_irq(struct cas *cp, const int ring) { if (ring == 0) { /* all but TX_DONE */ @@ -497,7 +473,6 @@ static int cas_page_free(struct cas *cp, cas_page_t *page) { pci_unmap_page(cp->pdev, page->dma_addr, cp->page_size, PCI_DMA_FROMDEVICE); - cas_buffer_dec(page); __free_pages(page->buffer, cp->page_order); kfree(page); return 0; @@ -527,7 +502,6 @@ static cas_page_t *cas_page_alloc(struct cas *cp, const gfp_t flags) page->buffer = alloc_pages(flags, cp->page_order); if (!page->buffer) goto page_err; - cas_buffer_init(page); page->dma_addr = pci_map_page(cp->pdev, page->buffer, 0, cp->page_size, PCI_DMA_FROMDEVICE); return page; @@ -558,8 +532,7 @@ static void cas_spare_free(struct cas *cp) /* free spare buffers */ INIT_LIST_HEAD(&list); spin_lock(&cp->rx_spare_lock); - list_splice(&cp->rx_spare_list, &list); - INIT_LIST_HEAD(&cp->rx_spare_list); + list_splice_init(&cp->rx_spare_list, &list); spin_unlock(&cp->rx_spare_lock); list_for_each_safe(elem, tmp, &list) { cas_page_free(cp, list_entry(elem, cas_page_t, list)); @@ -572,13 +545,11 @@ static void cas_spare_free(struct cas *cp) * lock than used everywhere else to manipulate this list. */ spin_lock(&cp->rx_inuse_lock); - list_splice(&cp->rx_inuse_list, &list); - INIT_LIST_HEAD(&cp->rx_inuse_list); + list_splice_init(&cp->rx_inuse_list, &list); spin_unlock(&cp->rx_inuse_lock); #else spin_lock(&cp->rx_spare_lock); - list_splice(&cp->rx_inuse_list, &list); - INIT_LIST_HEAD(&cp->rx_inuse_list); + list_splice_init(&cp->rx_inuse_list, &list); spin_unlock(&cp->rx_spare_lock); #endif list_for_each_safe(elem, tmp, &list) { @@ -599,14 +570,25 @@ static void cas_spare_recover(struct cas *cp, const gfp_t flags) /* make a local copy of the list */ INIT_LIST_HEAD(&list); spin_lock(&cp->rx_inuse_lock); - list_splice(&cp->rx_inuse_list, &list); - INIT_LIST_HEAD(&cp->rx_inuse_list); + list_splice_init(&cp->rx_inuse_list, &list); spin_unlock(&cp->rx_inuse_lock); list_for_each_safe(elem, tmp, &list) { cas_page_t *page = list_entry(elem, cas_page_t, list); - if (cas_buffer_count(page) > 1) + /* + * With the lockless pagecache, cassini buffering scheme gets + * slightly less accurate: we might find that a page has an + * elevated reference count here, due to a speculative ref, + * and skip it as in-use. Ideally we would be able to reclaim + * it. However this would be such a rare case, it doesn't + * matter too much as we should pick it up the next time round. + * + * Importantly, if we find that the page has a refcount of 1 + * here (our refcount), then we know it is definitely not inuse + * so we can reuse it. + */ + if (page_count(page->buffer) > 1) continue; list_del(elem); @@ -1374,7 +1356,7 @@ static inline cas_page_t *cas_page_spare(struct cas *cp, const int index) cas_page_t *page = cp->rx_pages[1][index]; cas_page_t *new; - if (cas_buffer_count(page) == 1) + if (page_count(page->buffer) == 1) return page; new = cas_page_dequeue(cp); @@ -1394,7 +1376,7 @@ static cas_page_t *cas_page_swap(struct cas *cp, const int ring, cas_page_t **page1 = cp->rx_pages[1]; /* swap if buffer is in use */ - if (cas_buffer_count(page0[index]) > 1) { + if (page_count(page0[index]->buffer) > 1) { cas_page_t *new = cas_page_spare(cp, index); if (new) { page1[index] = page0[index]; @@ -1979,6 +1961,7 @@ static int cas_rx_process_pkt(struct cas *cp, struct cas_rx_comp *rxc, struct cas_page *page; struct sk_buff *skb; void *addr, *crcaddr; + __sum16 csum; char *p; hlen = CAS_VAL(RX_COMP2_HDR_SIZE, words[1]); @@ -2062,10 +2045,10 @@ static int cas_rx_process_pkt(struct cas *cp, struct cas_rx_comp *rxc, skb_shinfo(skb)->nr_frags++; skb->data_len += hlen - swivel; + skb->truesize += hlen - swivel; skb->len += hlen - swivel; get_page(page->buffer); - cas_buffer_inc(page); frag->page = page->buffer; frag->page_offset = off; frag->size = hlen - swivel; @@ -2090,7 +2073,6 @@ static int cas_rx_process_pkt(struct cas *cp, struct cas_rx_comp *rxc, frag++; get_page(page->buffer); - cas_buffer_inc(page); frag->page = page->buffer; frag->page_offset = 0; frag->size = hlen; @@ -2158,16 +2140,20 @@ end_copy_pkt: skb_put(skb, alloclen); } - i = CAS_VAL(RX_COMP4_TCP_CSUM, words[3]); + csum = (__force __sum16)htons(CAS_VAL(RX_COMP4_TCP_CSUM, words[3])); if (cp->crc_size) { /* checksum includes FCS. strip it out. */ - i = csum_fold(csum_partial(crcaddr, cp->crc_size, i)); + csum = csum_fold(csum_partial(crcaddr, cp->crc_size, + csum_unfold(csum))); if (addr) cas_page_unmap(addr); } - skb->csum = ntohs(i ^ 0xffff); - skb->ip_summed = CHECKSUM_COMPLETE; skb->protocol = eth_type_trans(skb, cp->dev); + if (skb->protocol == htons(ETH_P_IP)) { + skb->csum = csum_unfold(~csum); + skb->ip_summed = CHECKSUM_COMPLETE; + } else + skb->ip_summed = CHECKSUM_NONE; return len; } @@ -2253,7 +2239,7 @@ static int cas_post_rxds_ringN(struct cas *cp, int ring, int num) released = 0; while (entry != last) { /* make a new buffer if it's still in use */ - if (cas_buffer_count(page[entry]) > 1) { + if (page_count(page[entry]->buffer) > 1) { cas_page_t *new = cas_page_dequeue(cp); if (!new) { /* let the timer know that we need to @@ -2611,7 +2597,7 @@ static int cas_poll(struct napi_struct *napi, int budget) { struct cas *cp = container_of(napi, struct cas, napi); struct net_device *dev = cp->dev; - int i, enable_intr, todo, credits; + int i, enable_intr, credits; u32 status = readl(cp->regs + REG_INTR_STATUS); unsigned long flags; @@ -4375,7 +4361,7 @@ static int cas_close(struct net_device *dev) struct cas *cp = netdev_priv(dev); #ifdef USE_NAPI - napi_enable(&cp->napi); + napi_disable(&cp->napi); #endif /* Make sure we don't get distracted by suspend/resume */ mutex_lock(&cp->pm_mutex); @@ -4419,7 +4405,7 @@ static struct { {"tx_fifo_errors"}, {"tx_packets"} }; -#define CAS_NUM_STAT_KEYS (sizeof(ethtool_cassini_statnames)/ETH_GSTRING_LEN) +#define CAS_NUM_STAT_KEYS ARRAY_SIZE(ethtool_cassini_statnames) static struct { const int offsets; /* neg. values for 2nd arg to cas_read_phy */ @@ -4872,6 +4858,90 @@ static int cas_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) return rc; } +/* When this chip sits underneath an Intel 31154 bridge, it is the + * only subordinate device and we can tweak the bridge settings to + * reflect that fact. + */ +static void __devinit cas_program_bridge(struct pci_dev *cas_pdev) +{ + struct pci_dev *pdev = cas_pdev->bus->self; + u32 val; + + if (!pdev) + return; + + if (pdev->vendor != 0x8086 || pdev->device != 0x537c) + return; + + /* Clear bit 10 (Bus Parking Control) in the Secondary + * Arbiter Control/Status Register which lives at offset + * 0x41. Using a 32-bit word read/modify/write at 0x40 + * is much simpler so that's how we do this. + */ + pci_read_config_dword(pdev, 0x40, &val); + val &= ~0x00040000; + pci_write_config_dword(pdev, 0x40, val); + + /* Max out the Multi-Transaction Timer settings since + * Cassini is the only device present. + * + * The register is 16-bit and lives at 0x50. When the + * settings are enabled, it extends the GRANT# signal + * for a requestor after a transaction is complete. This + * allows the next request to run without first needing + * to negotiate the GRANT# signal back. + * + * Bits 12:10 define the grant duration: + * + * 1 -- 16 clocks + * 2 -- 32 clocks + * 3 -- 64 clocks + * 4 -- 128 clocks + * 5 -- 256 clocks + * + * All other values are illegal. + * + * Bits 09:00 define which REQ/GNT signal pairs get the + * GRANT# signal treatment. We set them all. + */ + pci_write_config_word(pdev, 0x50, (5 << 10) | 0x3ff); + + /* The Read Prefecth Policy register is 16-bit and sits at + * offset 0x52. It enables a "smart" pre-fetch policy. We + * enable it and max out all of the settings since only one + * device is sitting underneath and thus bandwidth sharing is + * not an issue. + * + * The register has several 3 bit fields, which indicates a + * multiplier applied to the base amount of prefetching the + * chip would do. These fields are at: + * + * 15:13 --- ReRead Primary Bus + * 12:10 --- FirstRead Primary Bus + * 09:07 --- ReRead Secondary Bus + * 06:04 --- FirstRead Secondary Bus + * + * Bits 03:00 control which REQ/GNT pairs the prefetch settings + * get enabled on. Bit 3 is a grouped enabler which controls + * all of the REQ/GNT pairs from [8:3]. Bits 2 to 0 control + * the individual REQ/GNT pairs [2:0]. + */ + pci_write_config_word(pdev, 0x52, + (0x7 << 13) | + (0x7 << 10) | + (0x7 << 7) | + (0x7 << 4) | + (0xf << 0)); + + /* Force cacheline size to 0x8 */ + pci_write_config_byte(pdev, PCI_CACHE_LINE_SIZE, 0x08); + + /* Force latency timer to maximum setting so Cassini can + * sit on the bus as long as it likes. + */ + pci_write_config_byte(pdev, PCI_LATENCY_TIMER, 0xff); +} + static int __devinit cas_init_one(struct pci_dev *pdev, const struct pci_device_id *ent) { @@ -4927,6 +4997,8 @@ static int __devinit cas_init_one(struct pci_dev *pdev, printk(KERN_WARNING PFX "Could not enable MWI for %s\n", pci_name(pdev)); + cas_program_bridge(pdev); + /* * On some architectures, the default cache line size set * by pci_try_set_mwi reduces perforamnce. We have to increase @@ -5024,7 +5096,7 @@ static int __devinit cas_init_one(struct pci_dev *pdev, /* give us access to cassini registers */ cp->regs = pci_iomap(pdev, 0, casreg_len); - if (cp->regs == 0UL) { + if (!cp->regs) { dev_err(&pdev->dev, "Cannot map device registers, aborting.\n"); goto err_out_free_res; }