From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ml01.01.org (Postfix) with ESMTPS id 04CA821DF9687 for ; Sun, 30 Jul 2017 07:04:50 -0700 (PDT) Received: from orsmga002.jf.intel.com ([10.7.209.21]) by orsmga102.jf.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 30 Jul 2017 07:06:57 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.40,436,1496127600"; d="scan'208";a="117382986" Received: from jiaxinwu-mobl2.ccr.corp.intel.com ([10.254.210.26]) by orsmga002.jf.intel.com with ESMTP; 30 Jul 2017 07:06:55 -0700 From: Jiaxin Wu To: edk2-devel@lists.01.org Cc: Hegde Nagaraj P , Subramanian Sriram , Ni Ruiyu , Ye Ting , Fu Siyuan , Wu Jiaxin Date: Sun, 30 Jul 2017 22:06:47 +0800 Message-Id: <1501423607-28280-1-git-send-email-jiaxin.wu@intel.com> X-Mailer: git-send-email 1.9.5.msysgit.1 Subject: [Patch] NetworkPkg/Ip6Dxe: Fix the IPv6 PXE boot option goes missing issue X-BeenThere: edk2-devel@lists.01.org X-Mailman-Version: 2.1.22 Precedence: list List-Id: EDK II Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 30 Jul 2017 14:04:50 -0000 This patch is to fix the potential issue recorded at Bugzilla 636: https://bugzilla.tianocore.org/show_bug.cgi?id=636 The issue is caused by the IPv6 policy switching after PXEv6 boot. When IP policy is changing, the IPv6 children used by PXE.UdpRead() will be destroyed. Then, PXE Stop() function is called to uninstall the devicePath protocol, which leads to the IPv6 PXE boot option goes missing. Through the above analysis, the IP driver should take the responsibility for the upper layer network stacks recovery by using ConnectController(). Cc: Hegde Nagaraj P Cc: Subramanian Sriram Cc: Ni Ruiyu Cc: Ye Ting Cc: Fu Siyuan Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Wu Jiaxin --- NetworkPkg/Ip6Dxe/Ip6ConfigImpl.c | 72 ++++++++++++++++++++++++++++----------- 1 file changed, 53 insertions(+), 19 deletions(-) diff --git a/NetworkPkg/Ip6Dxe/Ip6ConfigImpl.c b/NetworkPkg/Ip6Dxe/Ip6ConfigImpl.c index 7c7acc7..f170716 100644 --- a/NetworkPkg/Ip6Dxe/Ip6ConfigImpl.c +++ b/NetworkPkg/Ip6Dxe/Ip6ConfigImpl.c @@ -45,16 +45,21 @@ VOID Ip6ConfigOnPolicyChanged ( IN IP6_SERVICE *IpSb, IN EFI_IP6_CONFIG_POLICY NewPolicy ) { - LIST_ENTRY *Entry; - LIST_ENTRY *Entry2; - LIST_ENTRY *Next; - IP6_INTERFACE *IpIf; - IP6_DAD_ENTRY *DadEntry; - IP6_DELAY_JOIN_LIST *DelayNode; + LIST_ENTRY *Entry; + LIST_ENTRY *Entry2; + LIST_ENTRY *Next; + IP6_INTERFACE *IpIf; + IP6_DAD_ENTRY *DadEntry; + IP6_DELAY_JOIN_LIST *DelayNode; + IP6_ADDRESS_INFO *AddrInfo; + IP6_PROTOCOL *Instance; + BOOLEAN Recovery; + + Recovery = FALSE; // // Currently there are only two policies: Manual and Automatic. Regardless of // what transition is going on, i.e., Manual -> Automatic and Automatic -> // Manual, we have to free default router list, on-link prefix list, autonomous @@ -78,22 +83,52 @@ Ip6ConfigOnPolicyChanged ( IP6_LINK_LOCAL_PREFIX_LENGTH, &IpSb->LinkLocalAddr ); } - // - // All IPv6 children that use global unicast address as it's source address - // should be destryoed now. The survivers are those use the link-local address - // or the unspecified address as the source address. - // TODO: Conduct a check here. - Ip6RemoveAddr ( - IpSb, - &IpSb->DefaultInterface->AddressList, - &IpSb->DefaultInterface->AddressCount, - NULL, - 0 - ); + if (!IsListEmpty (&IpSb->DefaultInterface->AddressList) && IpSb->DefaultInterface->AddressCount > 0) { + // + // If any IPv6 children (Instance) in configured state and use global unicast address, it will be + // destroyed in Ip6RemoveAddr() function later. Then, the upper layer driver's Stop() function will be + // called, which may break the upper layer network stacks. So, the driver should take the responsibility + // for the recovery by using ConnectController() after Ip6RemoveAddr(). + // Here, just check whether need to recover the upper layer network stacks later. + // + NET_LIST_FOR_EACH (Entry, &IpSb->DefaultInterface->AddressList) { + AddrInfo = NET_LIST_USER_STRUCT_S (Entry, IP6_ADDRESS_INFO, Link, IP6_ADDR_INFO_SIGNATURE); + if (!IsListEmpty (&IpSb->Children)) { + NET_LIST_FOR_EACH (Entry2, &IpSb->Children) { + Instance = NET_LIST_USER_STRUCT_S (Entry2, IP6_PROTOCOL, Link, IP6_PROTOCOL_SIGNATURE); + if ((Instance->State == IP6_STATE_CONFIGED) && EFI_IP6_EQUAL (&Instance->ConfigData.StationAddress, &AddrInfo->Address)) { + Recovery = TRUE; + break; + } + } + } + } + + // + // All IPv6 children that use global unicast address as it's source address + // should be destroyed now. The survivers are those use the link-local address + // or the unspecified address as the source address. + // TODO: Conduct a check here. + Ip6RemoveAddr ( + IpSb, + &IpSb->DefaultInterface->AddressList, + &IpSb->DefaultInterface->AddressCount, + NULL, + 0 + ); + + if (IpSb->Controller != NULL && Recovery) { + // + // ConnectController() to recover the upper layer network stacks. + // + gBS->ConnectController (IpSb->Controller, NULL, NULL, TRUE); + } + } + NET_LIST_FOR_EACH (Entry, &IpSb->Interfaces) { // // remove all pending delay node and DAD entries for the global addresses. // @@ -128,11 +163,10 @@ Ip6ConfigOnPolicyChanged ( // // delay 1 second // IpSb->Ticks = (UINT32) IP6_GET_TICKS (IP6_ONE_SECOND_IN_MS); } - } /** The work function to trigger the DHCPv6 process to perform a stateful autoconfiguration. -- 1.9.5.msysgit.1