diff --git a/src/mail.cpp b/src/mail.cpp index fefe2f9..4bdf8a6 100644 --- a/src/mail.cpp +++ b/src/mail.cpp @@ -1,3071 +1,3095 @@ /* @file mail.h * @brief High level class to work with Outlook Mailitems. * * Copyright (C) 2015, 2016 by Bundesamt für Sicherheit in der Informationstechnik * Software engineering by Intevation GmbH * * This file is part of GpgOL. * * GpgOL is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * GpgOL is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public License * along with this program; if not, see . */ #include "config.h" #include "dialogs.h" #include "common.h" #include "mail.h" #include "eventsinks.h" #include "attachment.h" #include "mapihelp.h" #include "mimemaker.h" #include "revert.h" #include "gpgoladdin.h" #include "mymapitags.h" #include "parsecontroller.h" #include "cryptcontroller.h" #include "windowmessages.h" #include "mlang-charset.h" #include "wks-helper.h" #include "keycache.h" #include "cpphelp.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #undef _ #define _(a) utf8_gettext (a) using namespace GpgME; static std::map s_mail_map; static std::map s_uid_map; static std::set uids_searched; static Mail *s_last_mail; #define COPYBUFSIZE (8 * 1024) Mail::Mail (LPDISPATCH mailitem) : m_mailitem(mailitem), m_processed(false), m_needs_wipe(false), m_needs_save(false), m_crypt_successful(false), m_is_smime(false), m_is_smime_checked(false), m_is_signed(false), m_is_valid(false), m_close_triggered(false), m_is_html_alternative(false), m_needs_encrypt(false), m_moss_position(0), m_crypto_flags(0), m_cached_html_body(nullptr), m_cached_plain_body(nullptr), m_cached_recipients(nullptr), m_type(MSGTYPE_UNKNOWN), m_do_inline(false), m_is_gsuite(false), m_crypt_state(NoCryptMail), m_window(nullptr), m_async_crypt_disabled(false), m_is_forwarded_crypto_mail(false), m_is_send_again(false), m_disable_att_remove_warning(false) { if (get_mail_for_item (mailitem)) { log_error ("Mail object for item: %p already exists. Bug.", mailitem); return; } m_event_sink = install_MailItemEvents_sink (mailitem); if (!m_event_sink) { /* Should not happen but in that case we don't add us to the list and just release the Mail item. */ log_error ("%s:%s: Failed to install MailItemEvents sink.", SRCNAME, __func__); gpgol_release(mailitem); return; } s_mail_map.insert (std::pair (mailitem, this)); s_last_mail = this; } GPGRT_LOCK_DEFINE(dtor_lock); Mail::~Mail() { /* This should fix a race condition where the mail is deleted before the parser is accessed in the decrypt thread. The shared_ptr of the parser then ensures that the parser is alive even if the mail is deleted while parsing. */ gpgrt_lock_lock (&dtor_lock); log_oom_extra ("%s:%s: dtor: Mail: %p item: %p", SRCNAME, __func__, this, m_mailitem); std::map::iterator it; log_oom_extra ("%s:%s: Detaching event sink", SRCNAME, __func__); detach_MailItemEvents_sink (m_event_sink); gpgol_release(m_event_sink); log_oom_extra ("%s:%s: Erasing mail", SRCNAME, __func__); it = s_mail_map.find(m_mailitem); if (it != s_mail_map.end()) { s_mail_map.erase (it); } if (!m_uuid.empty()) { auto it2 = s_uid_map.find(m_uuid); if (it2 != s_uid_map.end()) { s_uid_map.erase (it2); } } log_oom_extra ("%s:%s: releasing mailitem", SRCNAME, __func__); gpgol_release(m_mailitem); xfree (m_cached_html_body); xfree (m_cached_plain_body); release_cArray (m_cached_recipients); if (!m_uuid.empty()) { log_oom_extra ("%s:%s: destroyed: %p uuid: %s", SRCNAME, __func__, this, m_uuid.c_str()); } else { log_oom_extra ("%s:%s: non crypto (or sent) mail: %p destroyed", SRCNAME, __func__, this); } log_oom_extra ("%s:%s: nulling shared pointer", SRCNAME, __func__); m_parser = nullptr; m_crypter = nullptr; gpgrt_lock_unlock (&dtor_lock); log_oom_extra ("%s:%s: returning", SRCNAME, __func__); } Mail * Mail::get_mail_for_item (LPDISPATCH mailitem) { if (!mailitem) { return NULL; } std::map::iterator it; it = s_mail_map.find(mailitem); if (it == s_mail_map.end()) { return NULL; } return it->second; } Mail * Mail::get_mail_for_uuid (const char *uuid) { if (!uuid) { return NULL; } auto it = s_uid_map.find(std::string(uuid)); if (it == s_uid_map.end()) { return NULL; } return it->second; } bool Mail::is_valid_ptr (const Mail *mail) { auto it = s_mail_map.begin(); while (it != s_mail_map.end()) { if (it->second == mail) return true; ++it; } return false; } int Mail::pre_process_message () { LPMESSAGE message = get_oom_base_message (m_mailitem); if (!message) { log_error ("%s:%s: Failed to get base message.", SRCNAME, __func__); return 0; } log_oom_extra ("%s:%s: GetBaseMessage OK for %p.", SRCNAME, __func__, m_mailitem); /* Change the message class here. It is important that we change the message class in the before read event regardless if it is already set to one of GpgOL's message classes. Changing the message class (even if we set it to the same value again that it already has) causes Outlook to reconsider what it "knows" about a message and reread data from the underlying base message. */ mapi_change_message_class (message, 1, &m_type); if (m_type == MSGTYPE_UNKNOWN) { gpgol_release (message); return 0; } /* Create moss attachments here so that they are properly hidden when the item is read into the model. */ m_moss_position = mapi_mark_or_create_moss_attach (message, m_type); if (!m_moss_position) { log_error ("%s:%s: Failed to find moss attachment.", SRCNAME, __func__); m_type = MSGTYPE_UNKNOWN; } gpgol_release (message); return 0; } static LPDISPATCH get_attachment (LPDISPATCH mailitem, int pos) { LPDISPATCH attachment; LPDISPATCH attachments = get_oom_object (mailitem, "Attachments"); if (!attachments) { log_debug ("%s:%s: Failed to get attachments.", SRCNAME, __func__); return NULL; } std::string item_str; int count = get_oom_int (attachments, "Count"); if (count < 1) { log_debug ("%s:%s: Invalid attachment count: %i.", SRCNAME, __func__, count); gpgol_release (attachments); return NULL; } if (pos > 0) { item_str = std::string("Item(") + std::to_string(pos) + ")"; } else { item_str = std::string("Item(") + std::to_string(count) + ")"; } attachment = get_oom_object (attachments, item_str.c_str()); gpgol_release (attachments); return attachment; } /** Helper to check that all attachments are hidden, to be called before crypto. */ int Mail::check_attachments () const { LPDISPATCH attachments = get_oom_object (m_mailitem, "Attachments"); if (!attachments) { log_debug ("%s:%s: Failed to get attachments.", SRCNAME, __func__); return 1; } int count = get_oom_int (attachments, "Count"); if (!count) { gpgol_release (attachments); return 0; } std::string message; if (is_encrypted () && is_signed ()) { message += _("Not all attachments were encrypted or signed.\n" "The unsigned / unencrypted attachments are:\n\n"); } else if (is_signed ()) { message += _("Not all attachments were signed.\n" "The unsigned attachments are:\n\n"); } else if (is_encrypted ()) { message += _("Not all attachments were encrypted.\n" "The unencrypted attachments are:\n\n"); } else { gpgol_release (attachments); return 0; } bool foundOne = false; for (int i = 1; i <= count; i++) { std::string item_str; item_str = std::string("Item(") + std::to_string (i) + ")"; LPDISPATCH oom_attach = get_oom_object (attachments, item_str.c_str ()); if (!oom_attach) { log_error ("%s:%s: Failed to get attachment.", SRCNAME, __func__); continue; } VARIANT var; VariantInit (&var); if (get_pa_variant (oom_attach, PR_ATTACHMENT_HIDDEN_DASL, &var) || (var.vt == VT_BOOL && var.boolVal == VARIANT_FALSE)) { foundOne = true; char *dispName = get_oom_string (oom_attach, "DisplayName"); message += dispName ? dispName : "Unknown"; xfree (dispName); message += "\n"; } VariantClear (&var); gpgol_release (oom_attach); } gpgol_release (attachments); if (foundOne) { message += "\n"; message += _("Note: The attachments may be encrypted or signed " "on a file level but the GpgOL status does not apply to them."); wchar_t *wmsg = utf8_to_wchar (message.c_str ()); wchar_t *wtitle = utf8_to_wchar (_("GpgOL Warning")); MessageBoxW (get_active_hwnd (), wmsg, wtitle, MB_ICONWARNING|MB_OK); xfree (wmsg); xfree (wtitle); } return 0; } /** Get the cipherstream of the mailitem. */ static LPSTREAM get_attachment_stream (LPDISPATCH mailitem, int pos) { if (!pos) { log_debug ("%s:%s: Called with zero pos.", SRCNAME, __func__); return NULL; } LPDISPATCH attachment = get_attachment (mailitem, pos); LPSTREAM stream = NULL; if (!attachment) { // For opened messages that have ms-tnef type we // create the moss attachment but don't find it // in the OOM. Try to find it through MAPI. HRESULT hr; log_debug ("%s:%s: Failed to find MOSS Attachment. " "Fallback to MAPI.", SRCNAME, __func__); LPMESSAGE message = get_oom_message (mailitem); if (!message) { log_debug ("%s:%s: Failed to get MAPI Interface.", SRCNAME, __func__); return NULL; } hr = message->OpenProperty (PR_BODY_A, &IID_IStream, 0, 0, (LPUNKNOWN*)&stream); if (hr) { log_debug ("%s:%s: OpenProperty failed: hr=%#lx", SRCNAME, __func__, hr); return NULL; } return stream; } LPATTACH mapi_attachment = NULL; mapi_attachment = (LPATTACH) get_oom_iunknown (attachment, "MapiObject"); gpgol_release (attachment); if (!mapi_attachment) { log_debug ("%s:%s: Failed to get MapiObject of attachment: %p", SRCNAME, __func__, attachment); return NULL; } if (FAILED (mapi_attachment->OpenProperty (PR_ATTACH_DATA_BIN, &IID_IStream, 0, MAPI_MODIFY, (LPUNKNOWN*) &stream))) { log_debug ("%s:%s: Failed to open stream for mapi_attachment: %p", SRCNAME, __func__, mapi_attachment); gpgol_release (mapi_attachment); } return stream; } #if 0 This should work. But Outlook says no. See the comment in set_pa_variant about this. I left the code here as an example how to work with safearrays and how this probably should work. static int copy_data_property(LPDISPATCH target, std::shared_ptr attach) { VARIANT var; VariantInit (&var); /* Get the size */ off_t size = attach->get_data ().seek (0, SEEK_END); attach->get_data ().seek (0, SEEK_SET); if (!size) { TRACEPOINT; return 1; } if (!get_pa_variant (target, PR_ATTACH_DATA_BIN_DASL, &var)) { log_debug("Have variant. type: %x", var.vt); } else { log_debug("failed to get variant."); } /* Set the type to an array of unsigned chars (OLE SAFEARRAY) */ var.vt = VT_ARRAY | VT_UI1; /* Set up the bounds structure */ SAFEARRAYBOUND rgsabound[1]; rgsabound[0].cElements = static_cast (size); rgsabound[0].lLbound = 0; /* Create an OLE SAFEARRAY */ var.parray = SafeArrayCreate (VT_UI1, 1, rgsabound); if (var.parray == NULL) { TRACEPOINT; VariantClear(&var); return 1; } void *buffer = NULL; /* Get a safe pointer to the array */ if (SafeArrayAccessData(var.parray, &buffer) != S_OK) { TRACEPOINT; VariantClear(&var); return 1; } /* Copy data to it */ size_t nread = attach->get_data ().read (buffer, static_cast (size)); if (nread != static_cast (size)) { TRACEPOINT; VariantClear(&var); return 1; } /*/ Unlock the variant data */ if (SafeArrayUnaccessData(var.parray) != S_OK) { TRACEPOINT; VariantClear(&var); return 1; } if (set_pa_variant (target, PR_ATTACH_DATA_BIN_DASL, &var)) { TRACEPOINT; VariantClear(&var); return 1; } VariantClear(&var); return 0; } #endif static int copy_attachment_to_file (std::shared_ptr att, HANDLE hFile) { char copybuf[COPYBUFSIZE]; size_t nread; /* Security considerations: Writing the data to a temporary file is necessary as neither MAPI manipulation works in the read event to transmit the data nor Property Accessor works (see above). From a security standpoint there is a short time where the temporary files are on disk. Tempdir should be protected so that only the user can read it. Thus we have a local attack that could also take the data out of Outlook. FILE_SHARE_READ is necessary so that outlook can read the file. A bigger concern is that the file is manipulated by another software to fake the signature state. So we keep the write exlusive to us. We delete the file before closing the write file handle. */ /* Make sure we start at the beginning */ att->get_data ().seek (0, SEEK_SET); while ((nread = att->get_data ().read (copybuf, COPYBUFSIZE))) { DWORD nwritten; if (!WriteFile (hFile, copybuf, nread, &nwritten, NULL)) { log_error ("%s:%s: Failed to write in tmp attachment.", SRCNAME, __func__); return 1; } if (nread != nwritten) { log_error ("%s:%s: Write truncated.", SRCNAME, __func__); return 1; } } return 0; } /** Sets some meta data on the last attachment added. The meta data is taken from the attachment object. */ static int fixup_last_attachment (LPDISPATCH mail, std::shared_ptr attachment) { /* Currently we only set content id */ if (attachment->get_content_id ().empty()) { log_debug ("%s:%s: Content id not found.", SRCNAME, __func__); return 0; } LPDISPATCH attach = get_attachment (mail, -1); if (!attach) { log_error ("%s:%s: No attachment.", SRCNAME, __func__); return 1; } int ret = put_pa_string (attach, PR_ATTACH_CONTENT_ID_DASL, attachment->get_content_id ().c_str()); gpgol_release (attach); return ret; } /** Helper to update the attachments of a mail object in oom. does not modify the underlying mapi structure. */ static int add_attachments(LPDISPATCH mail, std::vector > attachments) { int err = 0; for (auto att: attachments) { if (att->get_display_name().empty()) { log_error ("%s:%s: Ignoring attachment without display name.", SRCNAME, __func__); continue; } wchar_t* wchar_name = utf8_to_wchar (att->get_display_name().c_str()); HANDLE hFile; wchar_t* wchar_file = get_tmp_outfile (wchar_name, &hFile); if (copy_attachment_to_file (att, hFile)) { log_error ("%s:%s: Failed to copy attachment %s to temp file", SRCNAME, __func__, att->get_display_name().c_str()); err = 1; } if (add_oom_attachment (mail, wchar_file, wchar_name)) { log_error ("%s:%s: Failed to add attachment: %s", SRCNAME, __func__, att->get_display_name().c_str()); err = 1; } CloseHandle (hFile); if (!DeleteFileW (wchar_file)) { log_error ("%s:%s: Failed to delete tmp attachment for: %s", SRCNAME, __func__, att->get_display_name().c_str()); err = 1; } xfree (wchar_file); xfree (wchar_name); err = fixup_last_attachment (mail, att); } return err; } GPGRT_LOCK_DEFINE(parser_lock); static DWORD WINAPI do_parsing (LPVOID arg) { gpgrt_lock_lock (&dtor_lock); /* We lock with mail dtors so we can be sure the mail->parser call is valid. */ Mail *mail = (Mail *)arg; if (!Mail::is_valid_ptr (mail)) { log_debug ("%s:%s: canceling parsing for: %p already deleted", SRCNAME, __func__, arg); gpgrt_lock_unlock (&dtor_lock); return 0; } /* This takes a shared ptr of parser. So the parser is still valid when the mail is deleted. */ auto parser = mail->parser(); gpgrt_lock_unlock (&dtor_lock); gpgrt_lock_lock (&parser_lock); gpgrt_lock_lock (&invalidate_lock); /* We lock the parser here to avoid too many decryption attempts if there are multiple mailobjects which might have already been deleted (e.g. by quick switches of the mailview.) Let's rather be a bit slower. */ log_debug ("%s:%s: preparing the parser for: %p", SRCNAME, __func__, arg); if (!Mail::is_valid_ptr (mail)) { log_debug ("%s:%s: cancel for: %p already deleted", SRCNAME, __func__, arg); gpgrt_lock_unlock (&invalidate_lock); gpgrt_lock_unlock (&parser_lock); return 0; } if (!parser) { log_error ("%s:%s: no parser found for mail: %p", SRCNAME, __func__, arg); gpgrt_lock_unlock (&invalidate_lock); gpgrt_lock_unlock (&parser_lock); return -1; } parser->parse(); do_in_ui_thread (PARSING_DONE, arg); gpgrt_lock_unlock (&invalidate_lock); gpgrt_lock_unlock (&parser_lock); return 0; } /* How encryption is done: There are two modes of encryption. Synchronous and Async. If async is used depends on the value of mail->async_crypt_disabled. Synchronous crypto: > Send Event < | State NoCryptMail Needs Crypto ? (get_gpgol_draft_info_flags != 0) -> No: Pass send -> unencrypted mail. -> Yes: mail->update_oom_data State = Mail::NeedsFirstAfterWrite check_inline_response invoke_oom_method (m_object, "Save", NULL); > Write Event < Pass because is_crypto_mail is false (not a decrypted mail) > AfterWrite Event < | State NeedsFirstAfterWrite State = NeedsActualCrypo encrypt_sign_start collect_input_data -> Check if Inline PGP should be used do_crypt -> Resolve keys / do crypto State = NeedsUpdateInMAPI update_crypt_mapi crypter->update_mail_mapi if (inline) (Meaning PGP/Inline) <-- do nothing. else build MSOXSMIME attachment and clear body / attachments. State = NeedsUpdateInOOM <- Back to Send Event update_crypt_oom -> Cleans body or sets PGP/Inline body. (inline_body_to_body) State = WantsSendMIME or WantsSendInline -> Saftey check "has_crypted_or_empty_body" -> If MIME Mail do the T3656 check. Send. State order for "inline_response" (sync) Mails. NoCryptMail NeedsFirstAfterWrite NeedsActualCrypto NeedsUpdateInMAPI NeedsUpdateInOOM WantsSendMIME (or inline for PGP Inline) -> Send. State order for async Mails NoCryptMail NeedsFirstAfterWrite NeedsActualCrypto -> Cancel Send. Windowmessages -> Crypto Done NeedsUpdateInOOM NeedsSecondAfterWrite trigger Save. NeedsUpdateInMAPI WantsSendMIME trigger Send. */ static DWORD WINAPI do_crypt (LPVOID arg) { gpgrt_lock_lock (&dtor_lock); /* We lock with mail dtors so we can be sure the mail->parser call is valid. */ Mail *mail = (Mail *)arg; if (!Mail::is_valid_ptr (mail)) { log_debug ("%s:%s: canceling crypt for: %p already deleted", SRCNAME, __func__, arg); gpgrt_lock_unlock (&dtor_lock); return 0; } if (mail->crypt_state() != Mail::NeedsActualCrypt) { log_debug ("%s:%s: invalid state %i", SRCNAME, __func__, mail->crypt_state ()); mail->set_window_enabled (true); gpgrt_lock_unlock (&dtor_lock); return -1; } /* This takes a shared ptr of crypter. So the crypter is still valid when the mail is deleted. */ auto crypter = mail->crypter(); gpgrt_lock_unlock (&dtor_lock); if (!crypter) { log_error ("%s:%s: no crypter found for mail: %p", SRCNAME, __func__, arg); gpgrt_lock_unlock (&parser_lock); mail->set_window_enabled (true); return -1; } GpgME::Error err; int rc = crypter->do_crypto(err); gpgrt_lock_lock (&dtor_lock); if (!Mail::is_valid_ptr (mail)) { log_debug ("%s:%s: aborting crypt for: %p already deleted", SRCNAME, __func__, arg); gpgrt_lock_unlock (&dtor_lock); return 0; } mail->set_window_enabled (true); if (rc == -1 || err) { mail->reset_crypter (); crypter = nullptr; if (err) { char *buf = nullptr; gpgrt_asprintf (&buf, _("Crypto operation failed:\n%s"), err.asString()); gpgol_message_box (mail->get_window(), buf, _("GpgOL"), MB_OK); xfree (buf); } else { gpgol_bug (mail->get_window (), ERR_CRYPT_RESOLVER_FAILED); } } if (rc || err.isCanceled()) { log_debug ("%s:%s: crypto failed for: %p with: %i err: %i", SRCNAME, __func__, arg, rc, err.code()); mail->set_crypt_state (Mail::NoCryptMail); mail->reset_crypter (); crypter = nullptr; gpgrt_lock_unlock (&dtor_lock); return rc; } if (!mail->async_crypt_disabled ()) { mail->set_crypt_state (Mail::NeedsUpdateInOOM); gpgrt_lock_unlock (&dtor_lock); // This deletes the Mail in Outlook 2010 do_in_ui_thread (CRYPTO_DONE, arg); log_debug ("%s:%s: UI thread finished for %p", SRCNAME, __func__, arg); } else { mail->set_crypt_state (Mail::NeedsUpdateInMAPI); mail->update_crypt_mapi (); if (mail->crypt_state () == Mail::WantsSendMIME) { // For sync crypto we need to switch this. mail->set_crypt_state (Mail::NeedsUpdateInOOM); } else { // A bug! log_debug ("%s:%s: Resetting crypter because of state mismatch. %p", SRCNAME, __func__, arg); crypter = nullptr; mail->reset_crypter (); } gpgrt_lock_unlock (&dtor_lock); } /* This works around a bug in pinentry that it might bring the wrong window to front. So after encryption / signing we bring outlook back to front. See GnuPG-Bug-Id: T3732 */ do_in_ui_thread_async (BRING_TO_FRONT, nullptr); log_debug ("%s:%s: crypto thread for %p finished", SRCNAME, __func__, arg); return 0; } bool Mail::is_crypto_mail() const { if (m_type == MSGTYPE_UNKNOWN || m_type == MSGTYPE_GPGOL || m_type == MSGTYPE_SMIME) { /* Not a message for us. */ return false; } return true; } int Mail::decrypt_verify() { if (!is_crypto_mail()) { log_debug ("%s:%s: Decrypt Verify for non crypto mail: %p.", SRCNAME, __func__, m_mailitem); return 0; } if (m_needs_wipe) { log_error ("%s:%s: Decrypt verify called for msg that needs wipe: %p", SRCNAME, __func__, m_mailitem); return 1; } set_uuid (); m_processed = true; /* Insert placeholder */ char *placeholder_buf; if (m_type == MSGTYPE_GPGOL_WKS_CONFIRMATION) { gpgrt_asprintf (&placeholder_buf, opt.prefer_html ? decrypt_template_html : decrypt_template, "OpenPGP", _("Pubkey directory confirmation"), _("This is a confirmation request to publish your Pubkey in the " "directory for your domain.\n\n" "

If you did not request to publish your Pubkey in your providers " "directory, simply ignore this message.

\n")); } else if (gpgrt_asprintf (&placeholder_buf, opt.prefer_html ? decrypt_template_html : decrypt_template, is_smime() ? "S/MIME" : "OpenPGP", _("message"), _("Please wait while the message is being decrypted / verified...")) == -1) { log_error ("%s:%s: Failed to format placeholder.", SRCNAME, __func__); return 1; } if (opt.prefer_html) { m_orig_body = get_oom_string (m_mailitem, "HTMLBody"); if (put_oom_string (m_mailitem, "HTMLBody", placeholder_buf)) { log_error ("%s:%s: Failed to modify html body of item.", SRCNAME, __func__); } } else { m_orig_body = get_oom_string (m_mailitem, "Body"); if (put_oom_string (m_mailitem, "Body", placeholder_buf)) { log_error ("%s:%s: Failed to modify body of item.", SRCNAME, __func__); } } xfree (placeholder_buf); /* Do the actual parsing */ auto cipherstream = get_attachment_stream (m_mailitem, m_moss_position); if (m_type == MSGTYPE_GPGOL_WKS_CONFIRMATION) { WKSHelper::instance ()->handle_confirmation_read (this, cipherstream); return 0; } if (!cipherstream) { log_debug ("%s:%s: Failed to get cipherstream.", SRCNAME, __func__); return 1; } m_parser = std::shared_ptr (new ParseController (cipherstream, m_type)); m_parser->setSender(GpgME::UserID::addrSpecFromString(get_sender().c_str())); log_mime_parser ("%s:%s: Parser for \"%s\" is %p", SRCNAME, __func__, get_subject ().c_str(), m_parser.get()); gpgol_release (cipherstream); HANDLE parser_thread = CreateThread (NULL, 0, do_parsing, (LPVOID) this, 0, NULL); if (!parser_thread) { log_error ("%s:%s: Failed to create decrypt / verify thread.", SRCNAME, __func__); } CloseHandle (parser_thread); return 0; } void find_and_replace(std::string& source, const std::string &find, const std::string &replace) { for(std::string::size_type i = 0; (i = source.find(find, i)) != std::string::npos;) { source.replace(i, find.length(), replace); i += replace.length(); } } void Mail::update_body() { if (!m_parser) { TRACEPOINT; return; } const auto error = m_parser->get_formatted_error (); if (!error.empty()) { if (opt.prefer_html) { if (put_oom_string (m_mailitem, "HTMLBody", error.c_str ())) { log_error ("%s:%s: Failed to modify html body of item.", SRCNAME, __func__); } else { log_debug ("%s:%s: Set error html to: '%s'", SRCNAME, __func__, error.c_str ()); } } else { if (put_oom_string (m_mailitem, "Body", error.c_str ())) { log_error ("%s:%s: Failed to modify html body of item.", SRCNAME, __func__); } else { log_debug ("%s:%s: Set error plain to: '%s'", SRCNAME, __func__, error.c_str ()); } } return; } if (m_verify_result.error()) { log_error ("%s:%s: Verification failed. Restoring Body.", SRCNAME, __func__); if (opt.prefer_html) { if (put_oom_string (m_mailitem, "HTMLBody", m_orig_body.c_str ())) { log_error ("%s:%s: Failed to modify html body of item.", SRCNAME, __func__); } } else { if (put_oom_string (m_mailitem, "Body", m_orig_body.c_str ())) { log_error ("%s:%s: Failed to modify html body of item.", SRCNAME, __func__); } } return; } // No need to carry body anymore m_orig_body = std::string(); auto html = m_parser->get_html_body (); /** Outlook does not show newlines if \r\r\n is a newline. We replace these as apparently some other buggy MUA sends this. */ find_and_replace (html, "\r\r\n", "\r\n"); if (opt.prefer_html && !html.empty() && !m_block_html) { - char *converted = ansi_charset_to_utf8 (m_parser->get_html_charset().c_str(), - html.c_str(), html.size()); + const auto charset = m_parser->get_html_charset(); + + int codepage = 0; + if (html_charset.empty()) + { + codepage = get_oom_int (m_mailitem, "InternetCodepage"); + log_debug ("%s:%s: Did not find html charset." + " Using internet Codepage %i.", + SRCNAME, __func__, codepage); + } + + char *converted = ansi_charset_to_utf8 (charset.c_str(), html.c_str(), + html.size(), codepage); int ret = put_oom_string (m_mailitem, "HTMLBody", converted ? converted : ""); xfree (converted); if (ret) { log_error ("%s:%s: Failed to modify html body of item.", SRCNAME, __func__); } return; } auto body = m_parser->get_body (); if (body.empty () && m_block_html && !html.empty()) { #if 0 Sadly the following code still offers to load external references it might also be too dangerous if Outlook somehow autoloads the references as soon as the Body is put into HTML // Fallback to show HTML as plaintext if HTML display // is blocked. log_error ("%s:%s: No text body. Putting HTML into plaintext.", SRCNAME, __func__); char *converted = ansi_charset_to_utf8 (m_parser->get_html_charset().c_str(), html.c_str(), html.size()); int ret = put_oom_string (m_mailitem, "HTMLBody", converted ? converted : ""); xfree (converted); if (ret) { log_error ("%s:%s: Failed to modify html body of item.", SRCNAME, __func__); body = html; } else { char *plainBody = get_oom_string (m_mailitem, "Body"); if (!plainBody) { log_error ("%s:%s: Failed to obtain converted plain body.", SRCNAME, __func__); body = html; } else { ret = put_oom_string (m_mailitem, "HTMLBody", plainBody); xfree (plainBody); if (ret) { log_error ("%s:%s: Failed to put plain into html body of item.", SRCNAME, __func__); body = html; } else { return; } } } #endif body = html; std::string buf = _("HTML display disabled."); buf += "\n\n"; buf += _("For security reasons HTML content in unsigned, encrypted\n" "S/MIME mails cannot be displayed.\n\n" "Please ask the sender to sign the message or to send it as plain text."); gpgol_message_box (get_window(), buf.c_str() , _("GpgOL"), MB_OK); } find_and_replace (body, "\r\r\n", "\r\n"); - char *converted = ansi_charset_to_utf8 (m_parser->get_body_charset().c_str(), - body.c_str(), body.size()); + + const auto plain_charset = m_parser->get_body_charset(); + + int codepage = 0; + if (plain_charset.empty()) + { + codepage = get_oom_int (m_mailitem, "InternetCodepage"); + log_debug ("%s:%s: Did not find body charset. " + "Using internet Codepage %i.", + SRCNAME, __func__, codepage); + } + + char *converted = ansi_charset_to_utf8 (plain_charset.c_str(), + body.c_str(), body.size(), + codepage); int ret = put_oom_string (m_mailitem, "Body", converted ? converted : ""); xfree (converted); if (ret) { log_error ("%s:%s: Failed to modify body of item.", SRCNAME, __func__); } return; } void Mail::parsing_done() { TRACEPOINT; log_oom_extra ("Mail %p Parsing done for parser: %p", this, m_parser.get()); if (!m_parser) { /* This should not happen but it happens when outlook sends multiple ItemLoad events for the same Mail Object. In that case it could happen that one parser was already done while a second is now returning for the wrong mail (as it's looked up by uuid.) We have a check in get_uuid that the uuid was not in the map before (and the parser is replaced). So this really really should not happen. We handle it anyway as we crash otherwise. It should not happen because the parser is only created in decrypt_verify which is called in the read event. And even in there we check if the parser was set. */ log_error ("%s:%s: No parser obj. For mail: %p", SRCNAME, __func__, this); return; } /* Store the results. */ m_decrypt_result = m_parser->decrypt_result (); m_verify_result = m_parser->verify_result (); m_crypto_flags = 0; if (!m_decrypt_result.isNull()) { m_crypto_flags |= 1; } if (m_verify_result.numSignatures()) { m_crypto_flags |= 2; } update_sigstate (); m_needs_wipe = !m_is_send_again; TRACEPOINT; /* Set categories according to the result. */ update_categories (); TRACEPOINT; m_block_html = m_parser->shouldBlockHtml (); if (m_block_html) { // Just to be careful. set_block_status (); } TRACEPOINT; /* Update the body */ update_body(); TRACEPOINT; /* Check that there are no unsigned / unencrypted messages. */ check_attachments (); /* Update attachments */ if (add_attachments (m_mailitem, m_parser->get_attachments())) { log_error ("%s:%s: Failed to update attachments.", SRCNAME, __func__); } if (m_is_send_again) { log_debug ("%s:%s: I think that this is the send again of a crypto mail.", SRCNAME, __func__); /* We no longer want to be treated like a crypto mail. */ m_type = MSGTYPE_UNKNOWN; LPMESSAGE msg = get_oom_base_message (m_mailitem); if (!msg) { TRACEPOINT; } else { set_gpgol_draft_info_flags (msg, m_crypto_flags); gpgol_release (msg); } remove_our_attachments (); } log_debug ("%s:%s: Delayed invalidate to update sigstate.", SRCNAME, __func__); CloseHandle(CreateThread (NULL, 0, delayed_invalidate_ui, (LPVOID) this, 0, NULL)); TRACEPOINT; return; } int Mail::encrypt_sign_start () { if (m_crypt_state != NeedsActualCrypt) { log_debug ("%s:%s: invalid state %i", SRCNAME, __func__, m_crypt_state); return -1; } int flags = 0; if (!needs_crypto()) { return 0; } LPMESSAGE message = get_oom_base_message (m_mailitem); if (!message) { log_error ("%s:%s: Failed to get base message.", SRCNAME, __func__); return -1; } flags = get_gpgol_draft_info_flags (message); gpgol_release (message); const auto window = get_active_hwnd (); if (m_is_gsuite) { auto att_table = mapi_create_attach_table (message, 0); int n_att_usable = count_usable_attachments (att_table); mapi_release_attach_table (att_table); /* Check for attachments if we have some abort. */ if (n_att_usable) { wchar_t *w_title = utf8_to_wchar (_( "GpgOL: Oops, G Suite Sync account detected")); wchar_t *msg = utf8_to_wchar ( _("G Suite Sync breaks outgoing crypto mails " "with attachments.\nUsing crypto and attachments " "with G Suite Sync is not supported.\n\n" "See: https://dev.gnupg.org/T3545 for details.")); MessageBoxW (window, msg, w_title, MB_ICONINFORMATION|MB_OK); xfree (msg); xfree (w_title); return -1; } } m_do_inline = m_is_gsuite ? true : opt.inline_pgp; GpgME::Protocol proto = opt.enable_smime ? GpgME::UnknownProtocol: GpgME::OpenPGP; m_crypter = std::shared_ptr (new CryptController (this, flags & 1, flags & 2, proto)); // Careful from here on we have to check every // error condition with window enabling again. set_window_enabled (false); if (m_crypter->collect_data ()) { log_error ("%s:%s: Crypter for mail %p failed to collect data.", SRCNAME, __func__, this); set_window_enabled (true); return -1; } if (!m_async_crypt_disabled) { CloseHandle(CreateThread (NULL, 0, do_crypt, (LPVOID) this, 0, NULL)); } else { do_crypt (this); } return 0; } int Mail::needs_crypto () { LPMESSAGE message = get_oom_message (m_mailitem); bool ret; if (!message) { log_error ("%s:%s: Failed to get message.", SRCNAME, __func__); return false; } ret = get_gpgol_draft_info_flags (message); gpgol_release(message); return ret; } int Mail::wipe (bool force) { if (!m_needs_wipe && !force) { return 0; } log_debug ("%s:%s: Removing plaintext from mailitem: %p.", SRCNAME, __func__, m_mailitem); if (put_oom_string (m_mailitem, "HTMLBody", "")) { if (put_oom_string (m_mailitem, "Body", "")) { log_debug ("%s:%s: Failed to wipe mailitem: %p.", SRCNAME, __func__, m_mailitem); return -1; } return -1; } else { put_oom_string (m_mailitem, "Body", ""); } m_needs_wipe = false; return 0; } int Mail::update_oom_data () { char *buf = nullptr; log_debug ("%s:%s", SRCNAME, __func__); if (!is_crypto_mail()) { /* Update the body format. */ m_is_html_alternative = get_oom_int (m_mailitem, "BodyFormat") > 1; /* Store the body. It was not obvious for me (aheinecke) how to access this through MAPI. */ if (m_is_html_alternative) { log_debug ("%s:%s: Is html alternative mail.", SRCNAME, __func__); xfree (m_cached_html_body); m_cached_html_body = get_oom_string (m_mailitem, "HTMLBody"); } xfree (m_cached_plain_body); m_cached_plain_body = get_oom_string (m_mailitem, "Body"); release_cArray (m_cached_recipients); m_cached_recipients = get_recipients (); } /* For some reason outlook may store the recipient address in the send using account field. If we have SMTP we prefer the SenderEmailAddress string. */ if (is_crypto_mail ()) { /* This is the case where we are reading a mail and not composing. When composing we need to use the SendUsingAccount because if you send from the folder of userA but change the from to userB outlook will keep the SenderEmailAddress of UserA. This is all so horrible. */ buf = get_sender_SenderEMailAddress (m_mailitem); if (!buf) { /* Try the sender Object */ buf = get_sender_Sender (m_mailitem); } } if (!buf) { buf = get_sender_SendUsingAccount (m_mailitem, &m_is_gsuite); } if (!buf && !is_crypto_mail ()) { /* Try the sender Object */ buf = get_sender_Sender (m_mailitem); } if (!buf) { /* We don't have s sender object or SendUsingAccount, well, in that case fall back to the current user. */ buf = get_sender_CurrentUser (m_mailitem); } if (!buf) { log_debug ("%s:%s: All fallbacks failed.", SRCNAME, __func__); return -1; } m_sender = buf; xfree (buf); return 0; } std::string Mail::get_sender () { if (m_sender.empty()) update_oom_data(); return m_sender; } std::string Mail::get_cached_sender () { return m_sender; } int Mail::close_all_mails () { int err = 0; std::map::iterator it; TRACEPOINT; std::map mail_map_copy = s_mail_map; for (it = mail_map_copy.begin(); it != mail_map_copy.end(); ++it) { /* XXX For non racy code the is_valid_ptr check should not be necessary but we crashed sometimes closing a destroyed mail. */ if (!is_valid_ptr (it->second)) { log_debug ("%s:%s: Already deleted mail for %p", SRCNAME, __func__, it->first); continue; } if (!it->second->is_crypto_mail()) { continue; } if (close_inspector (it->second) || close (it->second)) { log_error ("Failed to close mail: %p ", it->first); /* Should not happen */ if (is_valid_ptr (it->second) && it->second->revert()) { err++; } } } return err; } int Mail::revert_all_mails () { int err = 0; std::map::iterator it; for (it = s_mail_map.begin(); it != s_mail_map.end(); ++it) { if (it->second->revert ()) { log_error ("Failed to revert mail: %p ", it->first); err++; continue; } it->second->set_needs_save (true); if (!invoke_oom_method (it->first, "Save", NULL)) { log_error ("Failed to save reverted mail: %p ", it->second); err++; continue; } } return err; } int Mail::wipe_all_mails () { int err = 0; std::map::iterator it; for (it = s_mail_map.begin(); it != s_mail_map.end(); ++it) { if (it->second->wipe ()) { log_error ("Failed to wipe mail: %p ", it->first); err++; } } return err; } int Mail::revert () { int err = 0; if (!m_processed) { return 0; } m_disable_att_remove_warning = true; err = gpgol_mailitem_revert (m_mailitem); if (err == -1) { log_error ("%s:%s: Message revert failed falling back to wipe.", SRCNAME, __func__); return wipe (); } /* We need to reprocess the mail next time around. */ m_processed = false; m_needs_wipe = false; m_disable_att_remove_warning = false; return 0; } bool Mail::is_smime () { msgtype_t msgtype; LPMESSAGE message; if (m_is_smime_checked) { return m_is_smime; } message = get_oom_message (m_mailitem); if (!message) { log_error ("%s:%s: No message?", SRCNAME, __func__); return false; } msgtype = mapi_get_message_type (message); m_is_smime = msgtype == MSGTYPE_GPGOL_OPAQUE_ENCRYPTED || msgtype == MSGTYPE_GPGOL_OPAQUE_SIGNED; /* Check if it is an smime mail. Multipart signed can also be true. */ if (!m_is_smime && msgtype == MSGTYPE_GPGOL_MULTIPART_SIGNED) { char *proto; char *ct = mapi_get_message_content_type (message, &proto, NULL); if (ct && proto) { m_is_smime = (!strcmp (proto, "application/pkcs7-signature") || !strcmp (proto, "application/x-pkcs7-signature")); } else { log_error ("%s:%s: No protocol in multipart / signed mail.", SRCNAME, __func__); } xfree (proto); xfree (ct); } gpgol_release (message); m_is_smime_checked = true; return m_is_smime; } static std::string get_string (LPDISPATCH item, const char *str) { char *buf = get_oom_string (item, str); if (!buf) return std::string(); std::string ret = buf; xfree (buf); return ret; } std::string Mail::get_subject() const { return get_string (m_mailitem, "Subject"); } std::string Mail::get_body() const { return get_string (m_mailitem, "Body"); } std::string Mail::get_html_body() const { return get_string (m_mailitem, "HTMLBody"); } char ** Mail::get_recipients() const { LPDISPATCH recipients = get_oom_object (m_mailitem, "Recipients"); if (!recipients) { TRACEPOINT; return nullptr; } auto ret = get_oom_recipients (recipients); gpgol_release (recipients); return ret; } int Mail::close_inspector (Mail *mail) { LPDISPATCH inspector = get_oom_object (mail->item(), "GetInspector"); HRESULT hr; DISPID dispid; if (!inspector) { log_debug ("%s:%s: No inspector.", SRCNAME, __func__); return -1; } dispid = lookup_oom_dispid (inspector, "Close"); if (dispid != DISPID_UNKNOWN) { VARIANT aVariant[1]; DISPPARAMS dispparams; dispparams.rgvarg = aVariant; dispparams.rgvarg[0].vt = VT_INT; dispparams.rgvarg[0].intVal = 1; dispparams.cArgs = 1; dispparams.cNamedArgs = 0; hr = inspector->Invoke (dispid, IID_NULL, LOCALE_SYSTEM_DEFAULT, DISPATCH_METHOD, &dispparams, NULL, NULL, NULL); if (hr != S_OK) { log_debug ("%s:%s: Failed to close inspector: %#lx", SRCNAME, __func__, hr); gpgol_release (inspector); return -1; } } gpgol_release (inspector); return 0; } /* static */ int Mail::close (Mail *mail) { VARIANT aVariant[1]; DISPPARAMS dispparams; dispparams.rgvarg = aVariant; dispparams.rgvarg[0].vt = VT_INT; dispparams.rgvarg[0].intVal = 1; dispparams.cArgs = 1; dispparams.cNamedArgs = 0; log_oom_extra ("%s:%s: Invoking close for: %p", SRCNAME, __func__, mail->item()); mail->set_close_triggered (true); int rc = invoke_oom_method_with_parms (mail->item(), "Close", NULL, &dispparams); log_oom_extra ("%s:%s: Returned from close", SRCNAME, __func__); return rc; } void Mail::set_close_triggered (bool value) { m_close_triggered = value; } bool Mail::get_close_triggered () const { return m_close_triggered; } static const UserID get_uid_for_sender (const Key &k, const char *sender) { UserID ret; if (!sender) { return ret; } if (!k.numUserIDs()) { log_debug ("%s:%s: Key without uids", SRCNAME, __func__); return ret; } for (const auto uid: k.userIDs()) { if (!uid.email()) { log_error ("%s:%s: skipping uid without email.", SRCNAME, __func__); continue; } auto normalized_uid = uid.addrSpec(); auto normalized_sender = UserID::addrSpecFromString(sender); if (normalized_sender.empty() || normalized_uid.empty()) { log_error ("%s:%s: normalizing '%s' or '%s' failed.", SRCNAME, __func__, uid.email(), sender); continue; } if (normalized_sender == normalized_uid) { ret = uid; } } return ret; } void Mail::update_sigstate () { std::string sender = get_sender(); if (sender.empty()) { log_error ("%s:%s:%i", SRCNAME, __func__, __LINE__); return; } if (m_verify_result.isNull()) { log_debug ("%s:%s: No verify result.", SRCNAME, __func__); return; } if (m_verify_result.error()) { log_debug ("%s:%s: verify error.", SRCNAME, __func__); return; } for (const auto sig: m_verify_result.signatures()) { m_is_signed = true; m_uid = get_uid_for_sender (sig.key(), sender.c_str()); if (m_uid.isNull() || (sig.validity() != Signature::Validity::Marginal && sig.validity() != Signature::Validity::Full && sig.validity() != Signature::Validity::Ultimate)) { /* For our category we only care about trusted sigs. And the UID needs to match.*/ continue; } if (sig.validity() == Signature::Validity::Marginal) { const auto tofu = m_uid.tofuInfo(); if (!tofu.isNull() && (tofu.validity() != TofuInfo::Validity::BasicHistory && tofu.validity() != TofuInfo::Validity::LargeHistory)) { /* Marginal is only good enough without tofu. Otherwise we wait for basic trust. */ log_debug ("%s:%s: Discarding marginal signature." "With too little history.", SRCNAME, __func__); continue; } } log_debug ("%s:%s: Classified sender as verified uid validity: %i", SRCNAME, __func__, m_uid.validity()); m_sig = sig; m_is_valid = true; return; } log_debug ("%s:%s: No signature with enough trust. Using first", SRCNAME, __func__); m_sig = m_verify_result.signature(0); return; } bool Mail::is_valid_sig () { return m_is_valid; } void Mail::remove_categories () { const char *decCategory = _("GpgOL: Encrypted Message"); const char *verifyCategory = _("GpgOL: Trusted Sender Address"); remove_category (m_mailitem, decCategory); remove_category (m_mailitem, verifyCategory); } /* Now for some tasty hack: Outlook sometimes does not show the new categories properly but instead does some weird scrollbar thing. This can be avoided by resizing the message a bit. But somehow this only needs to be done once. Weird isn't it? But as this workaround worked let's do it programatically. Fun. Wan't some tomato sauce with this hack? */ static void resize_active_window () { HWND wnd = get_active_hwnd (); static std::vector resized_windows; if(std::find(resized_windows.begin(), resized_windows.end(), wnd) != resized_windows.end()) { /* We only need to do this once per window. XXX But sometimes we also need to do this once per view of the explorer. So for now this might break but we reduce the flicker. A better solution would be to find the current view and track that. */ return; } if (!wnd) { TRACEPOINT; return; } RECT oldpos; if (!GetWindowRect (wnd, &oldpos)) { TRACEPOINT; return; } if (!SetWindowPos (wnd, nullptr, (int)oldpos.left, (int)oldpos.top, /* Anything smaller then 19 was ignored when the window was * maximized on Windows 10 at least with a 1980*1024 * resolution. So I assume it's at least 1 percent. * This is all hackish and ugly but should work for 90%... * hopefully. */ (int)oldpos.right - oldpos.left - 20, (int)oldpos.bottom - oldpos.top, 0)) { TRACEPOINT; return; } if (!SetWindowPos (wnd, nullptr, (int)oldpos.left, (int)oldpos.top, (int)oldpos.right - oldpos.left, (int)oldpos.bottom - oldpos.top, 0)) { TRACEPOINT; return; } resized_windows.push_back(wnd); } void Mail::update_categories () { const char *decCategory = _("GpgOL: Encrypted Message"); const char *verifyCategory = _("GpgOL: Trusted Sender Address"); if (is_valid_sig()) { add_category (m_mailitem, verifyCategory); } else { remove_category (m_mailitem, verifyCategory); } if (!m_decrypt_result.isNull()) { add_category (m_mailitem, decCategory); } else { /* As a small safeguard against fakes we remove our categories */ remove_category (m_mailitem, decCategory); } resize_active_window(); return; } bool Mail::is_signed() const { return m_verify_result.numSignatures() > 0; } bool Mail::is_encrypted() const { return !m_decrypt_result.isNull(); } int Mail::set_uuid() { char *uuid; if (!m_uuid.empty()) { /* This codepath is reached by decrypt again after a close with discard changes. The close discarded the uuid on the OOM object so we have to set it again. */ log_debug ("%s:%s: Resetting uuid for %p to %s", SRCNAME, __func__, this, m_uuid.c_str()); uuid = get_unique_id (m_mailitem, 1, m_uuid.c_str()); } else { uuid = get_unique_id (m_mailitem, 1, nullptr); log_debug ("%s:%s: uuid for %p set to %s", SRCNAME, __func__, this, uuid); } if (!uuid) { log_debug ("%s:%s: Failed to get/set uuid for %p", SRCNAME, __func__, m_mailitem); return -1; } if (m_uuid.empty()) { m_uuid = uuid; Mail *other = get_mail_for_uuid (uuid); if (other) { /* According to documentation this should not happen as this means that multiple ItemLoad events occured for the same mailobject without unload / destruction of the mail. But it happens. If you invalidate the UI in the selection change event Outlook loads a new mailobject for the mail. Might happen in other surprising cases. We replace in that case as experiments have shown that the last mailobject is the one that is visible. Still troubling state so we log this as an error. */ log_error ("%s:%s: There is another mail for %p " "with uuid: %s replacing it.", SRCNAME, __func__, m_mailitem, uuid); delete other; } s_uid_map.insert (std::pair (m_uuid, this)); log_debug ("%s:%s: uuid for %p is now %s", SRCNAME, __func__, this, m_uuid.c_str()); } xfree (uuid); return 0; } /* Returns 2 if the userid is ultimately trusted. Returns 1 if the userid is fully trusted but has a signature by a key for which we have a secret and which is ultimately trusted. (Direct trust) 0 otherwise */ static int level_4_check (const UserID &uid) { if (uid.isNull()) { return 0; } if (uid.validity () == UserID::Validity::Ultimate) { return 2; } if (uid.validity () == UserID::Validity::Full) { for (const auto sig: uid.signatures ()) { const char *sigID = sig.signerKeyID (); if (sig.isNull() || !sigID) { /* should not happen */ TRACEPOINT; continue; } /* Direct trust information is not available through gnupg so we cached the keys with ultimate trust during parsing and now see if we find a direct trust path.*/ for (const auto secKey: ParseController::get_ultimate_keys ()) { /* Check that the Key id of the key matches */ const char *secKeyID = secKey.keyID (); if (!secKeyID || strcmp (secKeyID, sigID)) { continue; } /* Check that the userID of the signature is the ultimately trusted one. */ const char *sig_uid_str = sig.signerUserID(); if (!sig_uid_str) { /* should not happen */ TRACEPOINT; continue; } for (const auto signer_uid: secKey.userIDs ()) { if (signer_uid.validity() != UserID::Validity::Ultimate) { TRACEPOINT; continue; } const char *signer_uid_str = signer_uid.id (); if (!sig_uid_str) { /* should not happen */ TRACEPOINT; continue; } if (!strcmp(sig_uid_str, signer_uid_str)) { /* We have a match */ log_debug ("%s:%s: classified %s as ultimate because " "it was signed by uid %s of key %s", SRCNAME, __func__, signer_uid_str, sig_uid_str, secKeyID); return 1; } } } } } return 0; } std::string Mail::get_crypto_summary () { const int level = get_signature_level (); bool enc = is_encrypted (); if (level == 4 && enc) { return _("Security Level 4"); } if (level == 4) { return _("Trust Level 4"); } if (level == 3 && enc) { return _("Security Level 3"); } if (level == 3) { return _("Trust Level 3"); } if (level == 2 && enc) { return _("Security Level 2"); } if (level == 2) { return _("Trust Level 2"); } if (enc) { return _("Encrypted"); } if (is_signed ()) { /* Even if it is signed, if it is not validly signed it's still completly insecure as anyone could have signed this. So we avoid the label "signed" here as this word already implies some security. */ return _("Insecure"); } return _("Insecure"); } std::string Mail::get_crypto_one_line() { bool sig = is_signed (); bool enc = is_encrypted (); if (sig || enc) { if (sig && enc) { return _("Signed and encrypted message"); } else if (sig) { return _("Signed message"); } else if (enc) { return _("Encrypted message"); } } return _("Insecure message"); } std::string Mail::get_crypto_details() { std::string message; /* No signature with keys but error */ if (!is_encrypted() && !is_signed () && m_verify_result.error()) { message = _("You cannot be sure who sent, " "modified and read the message in transit."); message += "\n\n"; message += _("The message was signed but the verification failed with:"); message += "\n"; message += m_verify_result.error().asString(); return message; } /* No crypo, what are we doing here? */ if (!is_encrypted () && !is_signed ()) { return _("You cannot be sure who sent, " "modified and read the message in transit."); } /* Handle encrypt only */ if (is_encrypted() && !is_signed ()) { if (in_de_vs_mode ()) { if (m_sig.isDeVs()) { message += _("The encryption was VS-NfD-compliant."); } else { message += _("The encryption was not VS-NfD-compliant."); } } message += "\n\n"; message += _("You cannot be sure who sent the message because " "it is not signed."); return message; } bool keyFound = true; bool isOpenPGP = m_sig.key().isNull() ? !is_smime() : m_sig.key().protocol() == Protocol::OpenPGP; char *buf; bool hasConflict = false; int level = get_signature_level (); log_debug ("%s:%s: Formatting sig. Validity: %x Summary: %x Level: %i", SRCNAME, __func__, m_sig.validity(), m_sig.summary(), level); if (level == 4) { /* level 4 check for direct trust */ int four_check = level_4_check (m_uid); if (four_check == 2 && m_sig.key().hasSecret ()) { message = _("You signed this message."); } else if (four_check == 1) { message = _("The senders identity was certified by yourself."); } else if (four_check == 2) { message = _("The sender is allowed to certify identities for you."); } else { log_error ("%s:%s:%i BUG: Invalid sigstate.", SRCNAME, __func__, __LINE__); return message; } } else if (level == 3 && isOpenPGP) { /* Level three is only reachable through web of trust and no direct signature. */ message = _("The senders identity was certified by several trusted people."); } else if (level == 3 && !isOpenPGP) { /* Level three is the only level for trusted S/MIME keys. */ gpgrt_asprintf (&buf, _("The senders identity is certified by the trusted issuer:\n'%s'\n"), m_sig.key().issuerName()); message = buf; xfree (buf); } else if (level == 2 && m_uid.tofuInfo ().isNull ()) { /* Marginal trust through pgp only */ message = _("Some trusted people " "have certified the senders identity."); } else if (level == 2) { unsigned long first_contact = std::max (m_uid.tofuInfo().signFirst(), m_uid.tofuInfo().encrFirst()); char *time = format_date_from_gpgme (first_contact); /* i18n note signcount is always pulral because with signcount 1 we * would not be in this branch. */ gpgrt_asprintf (&buf, _("The senders address is trusted, because " "you have established a communication history " "with this address starting on %s.\n" "You encrypted %i and verified %i messages since."), time, m_uid.tofuInfo().encrCount(), m_uid.tofuInfo().signCount ()); xfree (time); message = buf; xfree (buf); } else if (level == 1) { /* This could be marginal trust through pgp, or tofu with little history. */ if (m_uid.tofuInfo ().signCount() == 1) { message += _("The senders signature was verified for the first time."); } else if (m_uid.tofuInfo ().validity() == TofuInfo::Validity::LittleHistory) { unsigned long first_contact = std::max (m_uid.tofuInfo().signFirst(), m_uid.tofuInfo().encrFirst()); char *time = format_date_from_gpgme (first_contact); gpgrt_asprintf (&buf, _("The senders address is not trustworthy yet because " "you only verified %i messages and encrypted %i messages to " "it since %s."), m_uid.tofuInfo().signCount (), m_uid.tofuInfo().encrCount (), time); xfree (time); message = buf; xfree (buf); } } else { /* Now we are in level 0, this could be a technical problem, no key or just unkown. */ message = is_encrypted () ? _("But the sender address is not trustworthy because:") : _("The sender address is not trustworthy because:"); message += "\n"; keyFound = !(m_sig.summary() & Signature::Summary::KeyMissing); bool general_problem = true; /* First the general stuff. */ if (m_sig.summary() & Signature::Summary::Red) { message += _("The signature is invalid: \n"); } else if (m_sig.summary() & Signature::Summary::SysError || m_verify_result.numSignatures() < 1) { message += _("There was an error verifying the signature.\n"); const auto err = m_sig.status (); if (err) { message += err.asString () + std::string ("\n"); } } else if (m_sig.summary() & Signature::Summary::SigExpired) { message += _("The signature is expired.\n"); } else { message += isOpenPGP ? _("The used key") : _("The used certificate"); message += " "; general_problem = false; } /* Now the key problems */ if ((m_sig.summary() & Signature::Summary::KeyMissing)) { message += _("is not available."); } else if ((m_sig.summary() & Signature::Summary::KeyRevoked)) { message += _("is revoked."); } else if ((m_sig.summary() & Signature::Summary::KeyExpired)) { message += _("is expired."); } else if ((m_sig.summary() & Signature::Summary::BadPolicy)) { message += _("is not meant for signing."); } else if ((m_sig.summary() & Signature::Summary::CrlMissing)) { message += _("could not be checked for revocation."); } else if ((m_sig.summary() & Signature::Summary::CrlTooOld)) { message += _("could not be checked for revocation."); } else if ((m_sig.summary() & Signature::Summary::TofuConflict) || m_uid.tofuInfo().validity() == TofuInfo::Conflict) { message += _("is not the same as the key that was used " "for this address in the past."); hasConflict = true; } else if (m_uid.isNull()) { gpgrt_asprintf (&buf, _("does not claim the address: \"%s\"."), get_sender().c_str()); message += buf; xfree (buf); } else if (((m_sig.validity() & Signature::Validity::Undefined) || (m_sig.validity() & Signature::Validity::Unknown) || (m_sig.summary() == Signature::Summary::None) || (m_sig.validity() == 0))&& !general_problem) { /* Bit of a catch all for weird results. */ if (isOpenPGP) { message += _("is not certified by any trustworthy key."); } else { message += _("is not certified by a trustworthy Certificate Authority or the Certificate Authority is unknown."); } } else if (m_uid.isRevoked()) { message += _("The sender marked this address as revoked."); } else if ((m_sig.validity() & Signature::Validity::Never)) { message += _("is marked as not trustworthy."); } } message += "\n\n"; if (in_de_vs_mode ()) { if (is_signed ()) { if (m_sig.isDeVs ()) { message += _("The signature is VS-NfD-compliant."); } else { message += _("The signature is not VS-NfD-compliant."); } message += "\n"; } if (is_encrypted ()) { if (m_decrypt_result.isDeVs ()) { message += _("The encryption is VS-NfD-compliant."); } else { message += _("The encryption is not VS-NfD-compliant."); } message += "\n\n"; } else { message += "\n"; } } if (hasConflict) { message += _("Click here to change the key used for this address."); } else if (keyFound) { message += isOpenPGP ? _("Click here for details about the key.") : _("Click here for details about the certificate."); } else { message += isOpenPGP ? _("Click here to search the key on the configured keyserver.") : _("Click here to search the certificate on the configured X509 keyserver."); } return message; } int Mail::get_signature_level () const { if (!m_is_signed) { return 0; } if (m_uid.isNull ()) { /* No m_uid matches our sender. */ return 0; } if (m_is_valid && (m_uid.validity () == UserID::Validity::Ultimate || (m_uid.validity () == UserID::Validity::Full && level_4_check (m_uid))) && (!in_de_vs_mode () || m_sig.isDeVs())) { return 4; } if (m_is_valid && m_uid.validity () == UserID::Validity::Full && (!in_de_vs_mode () || m_sig.isDeVs())) { return 3; } if (m_is_valid) { return 2; } if (m_sig.validity() == Signature::Validity::Marginal) { return 1; } if (m_sig.summary() & Signature::Summary::TofuConflict || m_uid.tofuInfo().validity() == TofuInfo::Conflict) { return 0; } return 0; } int Mail::get_crypto_icon_id () const { int level = get_signature_level (); int offset = is_encrypted () ? ENCRYPT_ICON_OFFSET : 0; return IDI_LEVEL_0 + level + offset; } const char* Mail::get_sig_fpr() const { if (!m_is_signed || m_sig.isNull()) { return nullptr; } return m_sig.fingerprint(); } /** Try to locate the keys for all recipients */ void Mail::locate_keys() { static bool locate_in_progress; if (locate_in_progress) { /** XXX The strangest thing seems to happen here: In get_recipients the lookup for "AddressEntry" on an unresolved address might cause network traffic. So Outlook somehow "detaches" this call and keeps processing window messages while the call is running. So our do_delayed_locate might trigger a second locate. If we access the OOM in this call while we access the same object in the blocked "detached" call we crash. (T3931) After the window message is handled outlook retunrs in the original lookup. A better fix here might be a non recursive lock of the OOM. But I expect that if we lock the handling of the Windowmessage we might deadlock. */ log_debug ("%s:%s: Locate for %p already in progress.", SRCNAME, __func__, this); return; } locate_in_progress = true; // First update oom data to have recipients and sender updated. update_oom_data (); char ** recipients = take_cached_recipients (); KeyCache::instance()->startLocateSecret (get_sender ().c_str ()); KeyCache::instance()->startLocate (get_sender ().c_str ()); KeyCache::instance()->startLocate (recipients); release_cArray (recipients); locate_in_progress = false; } bool Mail::is_html_alternative () const { return m_is_html_alternative; } char * Mail::take_cached_html_body () { char *ret = m_cached_html_body; m_cached_html_body = nullptr; return ret; } char * Mail::take_cached_plain_body () { char *ret = m_cached_plain_body; m_cached_plain_body = nullptr; return ret; } int Mail::get_crypto_flags () const { return m_crypto_flags; } void Mail::set_needs_encrypt (bool value) { m_needs_encrypt = value; } bool Mail::needs_encrypt() const { return m_needs_encrypt; } char ** Mail::take_cached_recipients() { char **ret = m_cached_recipients; m_cached_recipients = nullptr; return ret; } void Mail::append_to_inline_body (const std::string &data) { m_inline_body += data; } int Mail::inline_body_to_body() { if (!m_crypter) { log_error ("%s:%s: No crypter.", SRCNAME, __func__); return -1; } const auto body = m_crypter->get_inline_data (); if (body.empty()) { return 0; } int ret = put_oom_string (m_mailitem, "Body", body.c_str ()); return ret; } void Mail::update_crypt_mapi() { log_debug ("%s:%s: Update crypt mapi", SRCNAME, __func__); if (m_crypt_state != NeedsUpdateInMAPI) { log_debug ("%s:%s: invalid state %i", SRCNAME, __func__, m_crypt_state); return; } if (!m_crypter) { if (!m_mime_data.empty()) { log_debug ("%s:%s: Have override mime data creating dummy crypter", SRCNAME, __func__); m_crypter = std::shared_ptr (new CryptController (this, false, false, GpgME::UnknownProtocol)); } else { log_error ("%s:%s: No crypter.", SRCNAME, __func__); m_crypt_state = NoCryptMail; return; } } if (m_crypter->update_mail_mapi ()) { log_error ("%s:%s: Failed to update MAPI after crypt", SRCNAME, __func__); m_crypt_state = NoCryptMail; } else { m_crypt_state = WantsSendMIME; } /** If sync we need the crypter in update_crypt_oom */ if (!async_crypt_disabled ()) { // We don't need the crypter anymore. reset_crypter (); } } /** Checks in OOM if the body is either empty or contains the -----BEGIN tag. pair.first -> true if body starts with -----BEGIN pair.second -> true if body is empty. */ static std::pair has_crypt_or_empty_body_oom (Mail *mail) { auto body = mail->get_body(); std::pair ret; ret.first = false; ret.second = false; ltrim (body); if (body.size() > 10 && !strncmp (body.c_str(), "-----BEGIN", 10)) { ret.first = true; return ret; } if (!body.size()) { ret.second = true; } else { log_mime_parser ("%s:%s: Body found in %p : \"%s\"", SRCNAME, __func__, mail, body.c_str ()); } return ret; } void Mail::update_crypt_oom() { log_debug ("%s:%s: Update crypt oom for %p", SRCNAME, __func__, this); if (m_crypt_state != NeedsUpdateInOOM) { log_debug ("%s:%s: invalid state %i", SRCNAME, __func__, m_crypt_state); reset_crypter (); return; } if (do_pgp_inline ()) { if (inline_body_to_body ()) { log_error ("%s:%s: Inline body to body failed %p.", SRCNAME, __func__, this); gpgol_bug (get_active_hwnd(), ERR_INLINE_BODY_TO_BODY); m_crypt_state = NoCryptMail; return; } } if (m_crypter->get_protocol () == GpgME::CMS && m_crypter->is_encrypter ()) { /* We put the PIDNameContentType headers here for exchange because this is the only way we found to inject the smime-type. */ if (put_pa_string (m_mailitem, PR_PIDNameContentType_DASL, "application/pkcs7-mime;smime-type=\"enveloped-data\";name=smime.p7m")) { log_debug ("%s:%s: Failed to put PIDNameContentType for %p.", SRCNAME, __func__, this); } } /** When doing async update_crypt_mapi follows and needs the crypter. */ if (async_crypt_disabled ()) { reset_crypter (); } const auto pair = has_crypt_or_empty_body_oom (this); if (pair.first) { log_debug ("%s:%s: Looks like inline body. You can pass %p.", SRCNAME, __func__, this); m_crypt_state = WantsSendInline; return; } // We are in MIME land. Wipe the body. if (wipe (true)) { log_debug ("%s:%s: Cancel send for %p.", SRCNAME, __func__, this); wchar_t *title = utf8_to_wchar (_("GpgOL: Encryption not possible!")); wchar_t *msg = utf8_to_wchar (_( "Outlook returned an error when trying to send the encrypted mail.\n\n" "Please restart Outlook and try again.\n\n" "If it still fails consider using an encrypted attachment or\n" "switching to PGP/Inline in GpgOL's options.")); MessageBoxW (get_active_hwnd(), msg, title, MB_ICONERROR | MB_OK); xfree (msg); xfree (title); m_crypt_state = NoCryptMail; return; } m_crypt_state = NeedsSecondAfterWrite; return; } void Mail::set_window_enabled (bool value) { if (!value) { m_window = get_active_hwnd (); } log_debug ("%s:%s: enable window %p %i", SRCNAME, __func__, m_window, value); EnableWindow (m_window, value ? TRUE : FALSE); } bool Mail::check_inline_response () { /* Async sending might lead to crashes when the send invocation is done. * For now we treat every mail as an inline response to disable async * encryption. :-( For more details see: T3838 */ #ifdef DO_ASYNC_CRYPTO m_async_crypt_disabled = false; LPDISPATCH app = GpgolAddin::get_instance ()->get_application (); if (!app) { TRACEPOINT; return false; } LPDISPATCH explorer = get_oom_object (app, "ActiveExplorer"); if (!explorer) { TRACEPOINT; return false; } LPDISPATCH inlineResponse = get_oom_object (explorer, "ActiveInlineResponse"); gpgol_release (explorer); if (!inlineResponse) { return false; } // We have inline response // Check if we are it. It's a bit naive but meh. Worst case // is that we think inline response too often and do sync // crypt where we could do async crypt. char * inlineSubject = get_oom_string (inlineResponse, "Subject"); gpgol_release (inlineResponse); const auto subject = get_subject (); if (inlineResponse && !subject.empty() && !strcmp (subject.c_str (), inlineSubject)) { log_debug ("%s:%s: Detected inline response for '%p'", SRCNAME, __func__, this); m_async_crypt_disabled = true; } xfree (inlineSubject); #else m_async_crypt_disabled = true; #endif return m_async_crypt_disabled; } // static Mail * Mail::get_last_mail () { if (!s_last_mail || !is_valid_ptr (s_last_mail)) { s_last_mail = nullptr; } return s_last_mail; } // static void Mail::invalidate_last_mail () { s_last_mail = nullptr; } // static void Mail::locate_all_crypto_recipients() { if (!opt.autoresolve) { return; } std::map::iterator it; for (it = s_mail_map.begin(); it != s_mail_map.end(); ++it) { if (it->second->needs_crypto ()) { it->second->locate_keys (); } } } int Mail::remove_our_attachments () { LPDISPATCH attachments = get_oom_object (m_mailitem, "Attachments"); if (!attachments) { TRACEPOINT; return 0; } int count = get_oom_int (attachments, "Count"); LPDISPATCH to_delete[count]; int del_cnt = 0; for (int i = 1; i <= count; i++) { auto item_str = std::string("Item(") + std::to_string (i) + ")"; LPDISPATCH attachment = get_oom_object (attachments, item_str.c_str()); if (!attachment) { TRACEPOINT; continue; } attachtype_t att_type; if (get_pa_int (attachment, GPGOL_ATTACHTYPE_DASL, (int*) &att_type)) { /* Not our attachment. */ gpgol_release (attachment); continue; } if (att_type == ATTACHTYPE_PGPBODY || att_type == ATTACHTYPE_MOSS || att_type == ATTACHTYPE_MOSSTEMPL) { /* One of ours to delete. */ to_delete[del_cnt++] = attachment; /* Dont' release yet */ continue; } gpgol_release (attachment); } gpgol_release (attachments); int ret = 0; for (int i = 0; i < del_cnt; i++) { LPDISPATCH attachment = to_delete[i]; /* Delete the attachments that are marked to delete */ if (invoke_oom_method (attachment, "Delete", NULL)) { log_error ("%s:%s: Error: deleting attachment %i", SRCNAME, __func__, i); ret = -1; } gpgol_release (attachment); } return ret; } /* We are very verbose because if we fail it might mean that we have leaked plaintext -> critical. */ bool Mail::has_crypted_or_empty_body () { const auto pair = has_crypt_or_empty_body_oom (this); if (pair.first /* encrypted marker */) { log_debug ("%s:%s: Crypt Marker detected in OOM body. Return true %p.", SRCNAME, __func__, this); return true; } if (!pair.second) { log_debug ("%s:%s: Unexpected content detected. Return false %p.", SRCNAME, __func__, this); return false; } // Pair second == true (is empty) can happen on OOM error. LPMESSAGE message = get_oom_base_message (m_mailitem); if (!message && pair.second) { if (message) { gpgol_release (message); } return true; } size_t r_nbytes = 0; char *mapi_body = mapi_get_body (message, &r_nbytes); gpgol_release (message); if (!mapi_body || !r_nbytes) { // Body or bytes are null. we are empty. xfree (mapi_body); log_debug ("%s:%s: MAPI error or empty message. Return true. %p.", SRCNAME, __func__, this); return true; } if (r_nbytes > 10 && !strncmp (mapi_body, "-----BEGIN", 10)) { // Body is crypt. log_debug ("%s:%s: MAPI Crypt marker detected. Return true. %p.", SRCNAME, __func__, this); xfree (mapi_body); return true; } xfree (mapi_body); log_debug ("%s:%s: Found mapi body. Return false. %p.", SRCNAME, __func__, this); return false; } std::string Mail::get_verification_result_dump() { std::stringstream ss; ss << m_verify_result; return ss.str(); } void Mail::set_block_status() { SPropValue prop; LPMESSAGE message = get_oom_base_message (m_mailitem); prop.ulPropTag = PR_BLOCK_STATUS; prop.Value.l = 1; HRESULT hr = message->SetProps (1, &prop, NULL); if (hr) { log_error ("%s:%s: can't set block value: hr=%#lx\n", SRCNAME, __func__, hr); } gpgol_release (message); return; } void Mail::set_block_html(bool value) { m_block_html = value; } diff --git a/src/mlang-charset.cpp b/src/mlang-charset.cpp index d800893..934c1e3 100644 --- a/src/mlang-charset.cpp +++ b/src/mlang-charset.cpp @@ -1,117 +1,124 @@ /* @file mlang-charset.cpp * @brief Convert between charsets using Mlang * * Copyright (C) 2015 by Bundesamt für Sicherheit in der Informationstechnik * Software engineering by Intevation GmbH * * This file is part of GpgOL. * * GpgOL is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * GpgOL is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public License * along with this program; if not, see . */ #include "config.h" #include "common.h" #define INITGUID #include DEFINE_GUID (IID_IMultiLanguage, 0x275c23e1,0x3747,0x11d0,0x9f, 0xea,0x00,0xaa,0x00,0x3f,0x86,0x46); #include #undef INITGUID #include "mlang-charset.h" char *ansi_charset_to_utf8 (const char *charset, const char *input, - size_t inlen) + size_t inlen, int codepage) { LPMULTILANGUAGE multilang = NULL; MIMECSETINFO mime_info; HRESULT err; DWORD enc; DWORD mode = 0; unsigned int wlen = 0, uinlen = 0; wchar_t *buf; char *ret; - if (!charset || !strlen (charset)) + if ((!charset || !strlen (charset)) && !codepage) { - log_debug ("%s:%s: No charset returning plain.", + log_debug ("%s:%s: No charset / codepage returning plain.", SRCNAME, __func__); return strdup (input); } CoCreateInstance(CLSID_CMultiLanguage, NULL, CLSCTX_INPROC_SERVER, IID_IMultiLanguage, (void**)&multilang); if (!multilang) { log_error ("%s:%s: Failed to get multilang obj.", SRCNAME, __func__); return NULL; } if (inlen > UINT_MAX) { log_error ("%s:%s: Inlen too long. Bug.", SRCNAME, __func__); gpgol_release (multilang); return NULL; } uinlen = (unsigned int) inlen; - mime_info.uiCodePage = 0; - mime_info.uiInternetEncoding = 0; - BSTR w_charset = utf8_to_wchar (charset); - err = multilang->GetCharsetInfo (w_charset, &mime_info); - xfree (w_charset); - if (err != S_OK) + if (!codepage) { - log_error ("%s:%s: Failed to find charset for: %s", - SRCNAME, __func__, charset); - gpgol_release (multilang); - return strdup(input); + mime_info.uiCodePage = 0; + mime_info.uiInternetEncoding = 0; + BSTR w_charset = utf8_to_wchar (charset); + err = multilang->GetCharsetInfo (w_charset, &mime_info); + xfree (w_charset); + if (err != S_OK) + { + log_error ("%s:%s: Failed to find charset for: %s", + SRCNAME, __func__, charset); + gpgol_release (multilang); + return strdup(input); + } + enc = (mime_info.uiInternetEncoding == 0) ? mime_info.uiCodePage : + mime_info.uiInternetEncoding; + } + else + { + enc = codepage; } - enc = (mime_info.uiInternetEncoding == 0) ? mime_info.uiCodePage : - mime_info.uiInternetEncoding; /** Get the size of the result */ err = multilang->ConvertStringToUnicode(&mode, enc, const_cast(input), &uinlen, NULL, &wlen); if (FAILED (err)) { log_error ("%s:%s: Failed conversion.", SRCNAME, __func__); gpgol_release (multilang); return NULL; } buf = (wchar_t*) xmalloc(sizeof(wchar_t) * (wlen + 1)); err = multilang->ConvertStringToUnicode(&mode, enc, const_cast(input), &uinlen, buf, &wlen); gpgol_release (multilang); if (FAILED (err)) { log_error ("%s:%s: Failed conversion 2.", SRCNAME, __func__); xfree (buf); return NULL; } /* Doc is not clear if this is terminated. */ buf[wlen] = L'\0'; ret = wchar_to_utf8 (buf); xfree (buf); return ret; } diff --git a/src/mlang-charset.h b/src/mlang-charset.h index 06b0e68..7aaa26d 100644 --- a/src/mlang-charset.h +++ b/src/mlang-charset.h @@ -1,44 +1,45 @@ /* @file mlang-charset.h * @brief Convert between charsets using Mlang * * Copyright (C) 2015 by Bundesamt für Sicherheit in der Informationstechnik * Software engineering by Intevation GmbH * * This file is part of GpgOL. * * GpgOL is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * GpgOL is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public License * along with this program; if not, see . */ #include "common.h" #ifdef __cplusplus extern "C" { #if 0 } #endif #endif /** @brief convert input to utf8. * * @param charset: ANSI name of the charset to decode. * @param input: The input to convert. * @param inlen: The size of the input. + * @param codepage: Alternative codepage to be prefered over the name. * * @returns NULL on error or an UTF-8 encoded NULL terminated string. */ char *ansi_charset_to_utf8 (const char *charset, const char *input, - size_t inlen); + size_t inlen, int codepage); #ifdef __cplusplus } #endif diff --git a/src/rfc2047parse.c b/src/rfc2047parse.c index 17c7f73..5c96467 100644 --- a/src/rfc2047parse.c +++ b/src/rfc2047parse.c @@ -1,671 +1,671 @@ /* @file rfc2047parse.c * @brief Parsercode for rfc2047 * * Copyright (C) 2015 by Bundesamt für Sicherheit in der Informationstechnik * Software engineering by Intevation GmbH * * This file is part of GpgOL. * * GpgOL is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * GpgOL is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public License * along with this program; if not, see . */ /* This code is heavily based (mostly verbatim copy with glib * dependencies removed) on GMime rev 496313fb * modified by aheinecke@intevation.de * * Copyright (C) 2000-2014 Jeffrey Stedfast * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public License * as published by the Free Software Foundation; either version 2.1 * of the License, or (at your option) any later version. */ #ifdef HAVE_CONFIG_H #include #endif #include #include "common_indep.h" #include #ifdef HAVE_W32_SYSTEM # include "mlang-charset.h" #endif #include "gmime-table-private.h" /* mabye we need this at some point later? */ #define G_MIME_RFC2047_WORKAROUNDS 1 static unsigned char gmime_base64_rank[256] = { 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, 255,255,255,255,255,255,255,255,255,255,255, 62,255,255,255, 63, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61,255,255,255, 0,255,255, 255, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25,255,255,255,255,255, 255, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,255,255,255,255,255, 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, }; typedef struct _rfc2047_token { struct _rfc2047_token *next; char *charset; const char *text; size_t length; char encoding; char is_8bit; } rfc2047_token; static rfc2047_token * rfc2047_token_new (const char *text, size_t len) { rfc2047_token *token; token = xmalloc (sizeof (rfc2047_token)); memset (token, 0, sizeof (rfc2047_token)); token->length = len; token->text = text; return token; } static rfc2047_token * rfc2047_token_new_encoded_word (const char *word, size_t len) { rfc2047_token *token; const char *payload; char *charset; const char *inptr; const char *tmpchar; char *buf, *lang; char encoding; size_t n; /* check that this could even be an encoded-word token */ if (len < 7 || strncmp (word, "=?", 2) != 0 || strncmp (word + len - 2, "?=", 2) != 0) return NULL; /* skip over '=?' */ inptr = word + 2; tmpchar = inptr; if (*tmpchar == '?' || *tmpchar == '*') { /* this would result in an empty charset */ return NULL; } /* skip to the end of the charset */ if (!(inptr = memchr (inptr, '?', len - 2)) || inptr[2] != '?') return NULL; /* copy the charset into a buffer */ n = (size_t) (inptr - tmpchar); buf = malloc (n + 1); memcpy (buf, tmpchar, n); buf[n] = '\0'; charset = buf; /* rfc2231 updates rfc2047 encoded words... * The ABNF given in RFC 2047 for encoded-words is: * encoded-word := "=?" charset "?" encoding "?" encoded-text "?=" * This specification changes this ABNF to: * encoded-word := "=?" charset ["*" language] "?" encoding "?" encoded-text "?=" */ /* trim off the 'language' part if it's there... */ if ((lang = strchr (charset, '*'))) *lang = '\0'; /* skip over the '?' */ inptr++; /* make sure the first char after the encoding is another '?' */ if (inptr[1] != '?') return NULL; switch (*inptr++) { case 'B': case 'b': encoding = 'B'; break; case 'Q': case 'q': encoding = 'Q'; break; default: return NULL; } /* the payload begins right after the '?' */ payload = inptr + 1; /* find the end of the payload */ inptr = word + len - 2; /* make sure that we don't have something like: =?iso-8859-1?Q?= */ if (payload > inptr) return NULL; token = rfc2047_token_new (payload, inptr - payload); token->charset = charset; token->encoding = encoding; return token; } static void rfc2047_token_free (rfc2047_token * tok) { if (!tok) { return; } xfree (tok->charset); xfree (tok); } static rfc2047_token * tokenize_rfc2047_phrase (const char *in, size_t *len) { bool enable_rfc2047_workarounds = G_MIME_RFC2047_WORKAROUNDS; rfc2047_token list, *lwsp, *token, *tail; register const char *inptr = in; bool encoded = false; const char *text, *word; bool ascii; size_t n; tail = (rfc2047_token *) &list; list.next = NULL; lwsp = NULL; while (*inptr != '\0') { text = inptr; while (is_lwsp (*inptr)) inptr++; if (inptr > text) lwsp = rfc2047_token_new (text, inptr - text); else lwsp = NULL; word = inptr; ascii = true; if (is_atom (*inptr)) { if (enable_rfc2047_workarounds) { /* Make an extra effort to detect and * separate encoded-word tokens that * have been merged with other * words. */ if (!strncmp (inptr, "=?", 2)) { inptr += 2; /* skip past the charset (if one is even declared, sigh) */ while (*inptr && *inptr != '?') { ascii = ascii && is_ascii (*inptr); inptr++; } /* sanity check encoding type */ if (inptr[0] != '?' || !strchr ("BbQq", inptr[1]) || inptr[2] != '?') goto non_rfc2047; inptr += 3; /* find the end of the rfc2047 encoded word token */ while (*inptr && strncmp (inptr, "?=", 2) != 0) { ascii = ascii && is_ascii (*inptr); inptr++; } if (*inptr == '\0') { /* didn't find an end marker... */ inptr = word + 2; ascii = true; goto non_rfc2047; } inptr += 2; } else { non_rfc2047: /* stop if we encounter a possible rfc2047 encoded * token even if it's inside another word, sigh. */ while (is_atom (*inptr) && strncmp (inptr, "=?", 2) != 0) inptr++; } } else { while (is_atom (*inptr)) inptr++; } n = (size_t) (inptr - word); if ((token = rfc2047_token_new_encoded_word (word, n))) { /* rfc2047 states that you must ignore all * whitespace between encoded words */ if (!encoded && lwsp != NULL) { tail->next = lwsp; tail = lwsp; } else if (lwsp != NULL) { rfc2047_token_free (lwsp); } tail->next = token; tail = token; encoded = true; } else { /* append the lwsp and atom tokens */ if (lwsp != NULL) { tail->next = lwsp; tail = lwsp; } token = rfc2047_token_new (word, n); token->is_8bit = ascii ? 0 : 1; tail->next = token; tail = token; encoded = false; } } else { /* append the lwsp token */ if (lwsp != NULL) { tail->next = lwsp; tail = lwsp; } ascii = true; while (*inptr && !is_lwsp (*inptr) && !is_atom (*inptr)) { ascii = ascii && is_ascii (*inptr); inptr++; } n = (size_t) (inptr - word); token = rfc2047_token_new (word, n); token->is_8bit = ascii ? 0 : 1; tail->next = token; tail = token; encoded = false; } } *len = (size_t) (inptr - in); return list.next; } static void rfc2047_token_list_free (rfc2047_token * tokens) { rfc2047_token * cur = tokens; while (cur) { rfc2047_token *next = cur->next; rfc2047_token_free (cur); cur = next; } } /* this decodes rfc2047's version of quoted-printable */ static size_t quoted_decode (const unsigned char *in, size_t len, unsigned char *out, int *state, unsigned int *save) { register const unsigned char *inptr; register unsigned char *outptr; const unsigned char *inend; unsigned char c, c1; unsigned int saved; int need; if (len == 0) return 0; inend = in + len; outptr = out; inptr = in; need = *state; saved = *save; if (need > 0) { if (isxdigit ((int) *inptr)) { if (need == 1) { c = toupper ((int) (saved & 0xff)); c1 = toupper ((int) *inptr++); saved = 0; need = 0; goto decode; } saved = 0; need = 0; goto equals; } /* last encoded-word ended in a malformed quoted-printable sequence */ *outptr++ = '='; if (need == 1) *outptr++ = (char) (saved & 0xff); saved = 0; need = 0; } while (inptr < inend) { c = *inptr++; if (c == '=') { equals: if (inend - inptr >= 2) { if (isxdigit ((int) inptr[0]) && isxdigit ((int) inptr[1])) { c = toupper (*inptr++); c1 = toupper (*inptr++); decode: *outptr++ = (((c >= 'A' ? c - 'A' + 10 : c - '0') & 0x0f) << 4) | ((c1 >= 'A' ? c1 - 'A' + 10 : c1 - '0') & 0x0f); } else { /* malformed quoted-printable sequence? */ *outptr++ = '='; } } else { /* truncated payload, maybe it was split across encoded-words? */ if (inptr < inend) { if (isxdigit ((int) *inptr)) { saved = *inptr; need = 1; break; } else { /* malformed quoted-printable sequence? */ *outptr++ = '='; } } else { saved = 0; need = 2; break; } } } else if (c == '_') { /* _'s are an rfc2047 shortcut for encoding spaces */ *outptr++ = ' '; } else { *outptr++ = c; } } *state = need; *save = saved; return (size_t) (outptr - out); } /** * g_mime_encoding_base64_decode_step: * @inbuf: input buffer * @inlen: input buffer length * @outbuf: output buffer * @state: holds the number of bits that are stored in @save * @save: leftover bits that have not yet been decoded * * Decodes a chunk of base64 encoded data. * * Returns: the number of bytes decoded (which have been dumped in * @outbuf). **/ size_t g_mime_encoding_base64_decode_step (const unsigned char *inbuf, size_t inlen, unsigned char *outbuf, int *state, unsigned int *save) { register const unsigned char *inptr; register unsigned char *outptr; const unsigned char *inend; register unsigned int saved; unsigned char c; int npad, n, i; inend = inbuf + inlen; outptr = outbuf; inptr = inbuf; npad = (*state >> 8) & 0xff; n = *state & 0xff; saved = *save; /* convert 4 base64 bytes to 3 normal bytes */ while (inptr < inend) { c = gmime_base64_rank[*inptr++]; if (c != 0xff) { saved = (saved << 6) | c; n++; if (n == 4) { *outptr++ = saved >> 16; *outptr++ = saved >> 8; *outptr++ = saved; n = 0; if (npad > 0) { outptr -= npad; npad = 0; } } } } /* quickly scan back for '=' on the end somewhere */ /* fortunately we can drop 1 output char for each trailing '=' (up to 2) */ for (i = 2; inptr > inbuf && i; ) { inptr--; if (gmime_base64_rank[*inptr] != 0xff) { if (*inptr == '=' && outptr > outbuf) { if (n == 0) { /* we've got a complete quartet so it's safe to drop an output character. */ outptr--; } else if (npad < 2) { /* keep a record of the number of ='s at the end of the input stream, up to 2 */ npad++; } } i--; } } *state = (npad << 8) | n; *save = n ? saved : 0; return (outptr - outbuf); } static size_t rfc2047_token_decode (rfc2047_token *token, unsigned char *outbuf, int *state, unsigned int *save) { const unsigned char *inbuf = (const unsigned char *) token->text; size_t len = token->length; if (token->encoding == 'B') return g_mime_encoding_base64_decode_step (inbuf, len, outbuf, state, save); else return quoted_decode (inbuf, len, outbuf, state, save); } static char * rfc2047_decode_tokens (rfc2047_token *tokens, size_t buflen) { rfc2047_token *token, *next; size_t outlen, len, tmplen; unsigned char *outptr; const char *charset; char *outbuf; char *decoded; char encoding; unsigned int save; int state; char *str; decoded = xmalloc (buflen + 1); memset (decoded, 0, buflen + 1); tmplen = 76; outbuf = xmalloc (tmplen); token = tokens; while (token != NULL) { next = token->next; if (token->encoding) { /* In order to work around broken mailers, we need to combine * the raw decoded content of runs of identically encoded word * tokens before converting into UTF-8. */ encoding = token->encoding; charset = token->charset; len = token->length; state = 0; save = 0; /* find the end of the run (and measure the buffer length we'll need) */ while (next && next->encoding == encoding && !strcmp (next->charset, charset)) { len += next->length; next = next->next; } /* make sure our temporary output buffer is large enough... */ if (len > tmplen) { xrealloc (outbuf, len + 1); tmplen = len + 1; } /* base64 / quoted-printable decode each of the tokens... */ outptr = outbuf; outlen = 0; do { /* Note: by not resetting state/save each loop, we effectively * treat the payloads as one continuous block, thus allowing * us to handle cases where a hex-encoded triplet of a * quoted-printable encoded payload is split between 2 or more * encoded-word tokens. */ len = rfc2047_token_decode (token, outptr, &state, &save); token = token->next; outptr += len; outlen += len; } while (token != next); outptr = outbuf; /* convert the raw decoded text into UTF-8 */ if (!strcasecmp (charset, "UTF-8")) { strncat (decoded, (char *) outptr, outlen); } else { #ifdef HAVE_W32_SYSTEM - str = ansi_charset_to_utf8 (charset, outptr, outlen); + str = ansi_charset_to_utf8 (charset, outptr, outlen, 0); #else log_debug ("%s:%s: Conversion not available on non W32 systems", SRCNAME, __func__); str = strndup (outptr, outlen); #endif if (!str) { log_error ("%s:%s: Failed conversion from: %s for word: %s.", SRCNAME, __func__, charset, outptr); } else { strcat (decoded, str); xfree (str); } } } else { strncat (decoded, token->text, token->length); } if (token && token->is_8bit) { /* We don't support this. */ log_error ("%s:%s: Unknown 8bit encoding detected.", SRCNAME, __func__); } token = next; } xfree (outbuf); return decoded; } /** * g_mime_utils_header_decode_phrase: * @phrase: header to decode * * Decodes an rfc2047 encoded 'phrase' header. * * Note: See g_mime_set_user_charsets() for details on how charset * conversion is handled for unencoded 8bit text and/or wrongly * specified rfc2047 encoded-word tokens. * * Returns: a newly allocated UTF-8 string representing the the decoded * header. **/ static char * g_mime_utils_header_decode_phrase (const char *phrase) { rfc2047_token *tokens; char *decoded; size_t len; tokens = tokenize_rfc2047_phrase (phrase, &len); decoded = rfc2047_decode_tokens (tokens, len); rfc2047_token_list_free (tokens); return decoded; } /* Try to parse an rfc 2047 filename for attachment handling. returns the parsed string. On errors the input string is just copied with strdup */ char * rfc2047_parse (const char *input) { char *decoded; if (!input) return strdup (""); log_debug ("%s:%s: Input: \"%s\"", SRCNAME, __func__, input); decoded = g_mime_utils_header_decode_phrase (input); log_debug ("%s:%s: Decoded: \"%s\"", SRCNAME, __func__, decoded); if (!decoded || !strlen (decoded)) { xfree (decoded); return strdup (input); } return decoded; }