\nIt is common to have a critical section which protects against\nconcurrent writes to a variable or collection of variables.\nBut if you protect a write with a critical section,\nyou may also want to protect the read,\nbecause the read might race against the write.\n<\/p>\n
\n\nAdam Rosenfield<\/a>\nshared his experience with this issue\n\nin a comment from a few years back<\/a>.\nI’ll reproduce the example here in part to save you the trouble of clicking,\nbut also to make this entry look longer and consequently make it seem\nlike I’m actually doing some work (when in fact Adam did nearly all of the\nwork):\n<\/p>\n \nThere is a race condition here:\n<\/p>\n \nThe fix was to enclose the read<\/i> of \nForgetting to enclose the read inside a critical section is a common\noversight.\nI’ve made it myself more than once.\nYou say to yourself,\n“I don’t need a critical section here.\nI’m just reading a value which can safely be read atomically.”\nBut you forget that the critical section isn’t just for protecting\nthe write to the variable;\nit’s also to protect all the other actions that take place under\nthe critical section.\n<\/p>\n \nAnd just to make it so I actually did some work today,\nI leave you with this puzzle based on an actual customer problem:\n<\/p>\n \nThe real class was significantly more complicated than this,\nbut I’ve distilled the problem to its essence.\n<\/p>\n \nThe customer added,\n“I tried declaring It is common to have a critical section which protects against concurrent writes to a variable or collection of variables. But if you protect a write with a critical section, you may also want to protect the read, because the read might race against the write. Adam Rosenfield shared his experience with this issue in […]<\/p>\n","protected":false},"author":1069,"featured_media":111744,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[1],"tags":[25],"class_list":["post-9003","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-oldnewthing","tag-code"],"acf":[],"blog_post_summary":" It is common to have a critical section which protects against concurrent writes to a variable or collection of variables. But if you protect a write with a critical section, you may also want to protect the read, because the read might race against the write. Adam Rosenfield shared his experience with this issue in […]<\/p>\n","_links":{"self":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/posts\/9003","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/users\/1069"}],"replies":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/comments?post=9003"}],"version-history":[{"count":0,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/posts\/9003\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/media\/111744"}],"wp:attachment":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/media?parent=9003"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/categories?post=9003"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/tags?post=9003"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}\nclass X {\n volatile int mState;\n CRITICAL_SECTION mCrit;\n HANDLE mEvent;\n};\nX::Wait() {\n while(mState != kDone) {\n WaitForSingleObject(mEvent, INFINITE);\n }\n}\nX::~X() {\n DestroyCriticalSection(&mCrit);\n}\nX::SetState(int state) {\n EnterCriticalSection(&mCrit);\n \/\/ do some state logic\n mState = state;\n SetEvent(mEvent);\n LeaveCriticalSection(&mCrit);\n}\nThread1()\n{\n X x;\n ... spawn off thread 2 ...\n x.Wait();\n}\nThread2(X* px)\n{\n ...\n px->SetState(kDone);\n ...\n}<\/i>\n<\/pre>\n\n
X::Wait<\/code> and waits.\nX::SetState<\/code>.\nSetEvent<\/code>.<\/p>\nWaitForSingleObject<\/code>\n call, notices that mState == kDone<\/code>, and therefore\n returns from the X::Wait<\/code> method.<\/p>\nX<\/code> object, which\n destroys the critical section.<\/p>\nmState<\/code>\ninside a critical section:\n<\/p>\n\nX::Wait() {\n while(1) {\n EnterCriticalSection(&mCrit);\n int state = mState;\n LeaveCriticalSection(&mCrit);\n if(state == kDone)\n break;\n WaitForSingleObject(mEvent, INFINITE);\n }\n}\n<\/pre>\n\nclass BufferPool {\npublic:\n BufferPool() { ... }\n ~BufferPool() { ... }\n Buffer *GetBuffer()\n {\n Buffer *pBuffer = FindFreeBuffer();\n if (pBuffer) {\n pBuffer->mIsFree = false;\n }\n return pBuffer;\n }\n void ReturnBuffer(Buffer *pBuffer)\n {\n pBuffer->mIsFree = true;\n }\nprivate:\n Buffer *FindFreeBuffer()\n {\n EnterCriticalSection(&mCrit);\n Buffer *pBuffer = NULL;\n for (int i = 0; i < 8; i++) {\n if (mBuffers[i].mIsFree) {\n pBuffer = &mBuffers[i];\n break;\n }\n }\n LeaveCriticalSection(&mCrit);\n return pBuffer;\n }\nprivate:\n CRITICAL_SECTION mCrit;\n Buffer mBuffers[8];\n};<\/i>\n<\/pre>\nmIsFree<\/code> as a volatile variable,\nbut that didn’t seem to help.”<\/p>\n","protected":false},"excerpt":{"rendered":"